[FEATURE]: Add Snap packaging support with Github workflow

Fix SEI payload type handling: changes payload_type and payload_size from i32 to u32 for type safety, keeping as usize casts only where needed for indexing.

.github/workflows/build_deb.yml

@@ -0,0 +1,283 @@
+name: Build Linux .deb Package
+
+on:
+  # Build on releases
+  release:
+    types: [published]
+  # Allow manual trigger
+  workflow_dispatch:
+    inputs:
+      build_type:
+        description: 'Build type (all, basic, hardsubx)'
+        required: false
+        default: 'all'
+  # Build on pushes to workflow file for testing
+  push:
+    paths:
+      - '.github/workflows/build_deb.yml'
+
+jobs:
+  build-deb:
+    runs-on: ubuntu-24.04
+    strategy:
+      fail-fast: false
+      matrix:
+        build_type: [basic, hardsubx]
+
+    steps:
+      - name: Check if should build this variant
+        id: should_build
+        run: |
+          if [ "${{ github.event_name }}" = "workflow_dispatch" ]; then
+            INPUT_TYPE="${{ github.event.inputs.build_type }}"
+            if [ "$INPUT_TYPE" = "all" ] || [ "$INPUT_TYPE" = "${{ matrix.build_type }}" ]; then
+              echo "should_build=true" >> $GITHUB_OUTPUT
+            else
+              echo "should_build=false" >> $GITHUB_OUTPUT
+            fi
+          else
+            echo "should_build=true" >> $GITHUB_OUTPUT
+          fi
+
+      - name: Checkout repository
+        if: steps.should_build.outputs.should_build == 'true'
+        uses: actions/checkout@v6
+
+      - name: Get version
+        if: steps.should_build.outputs.should_build == 'true'
+        id: version
+        run: |
+          # Extract version from source or use tag
+          if [ "${{ github.event_name }}" = "release" ]; then
+            VERSION="${{ github.event.release.tag_name }}"
+            VERSION="${VERSION#v}"  # Remove 'v' prefix if present
+          else
+            # Extract version from lib_ccx.h (e.g., #define VERSION "0.96.5")
+            VERSION=$(grep -oP '#define VERSION "\K[^"]+' src/lib_ccx/lib_ccx.h || echo "0.96")
+          fi
+          echo "version=$VERSION" >> $GITHUB_OUTPUT
+          echo "Building version: $VERSION"
+
+      - name: Install base dependencies
+        if: steps.should_build.outputs.should_build == 'true'
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y --no-install-recommends \
+            build-essential \
+            cmake \
+            pkg-config \
+            zlib1g-dev \
+            libpng-dev \
+            libjpeg-dev \
+            libfreetype-dev \
+            libxml2-dev \
+            libcurl4-gnutls-dev \
+            libssl-dev \
+            clang \
+            libclang-dev \
+            tesseract-ocr \
+            libtesseract-dev \
+            libleptonica-dev \
+            patchelf
+
+      - name: Install FFmpeg dependencies (HardSubX)
+        if: steps.should_build.outputs.should_build == 'true' && matrix.build_type == 'hardsubx'
+        run: |
+          sudo apt-get install -y --no-install-recommends \
+            libavcodec-dev \
+            libavformat-dev \
+            libavutil-dev \
+            libswscale-dev \
+            libswresample-dev \
+            libavfilter-dev \
+            libavdevice-dev
+
+      - name: Install Rust toolchain
+        if: steps.should_build.outputs.should_build == 'true'
+        uses: dtolnay/rust-toolchain@stable
+
+      - name: Cache GPAC build
+        if: steps.should_build.outputs.should_build == 'true'
+        id: cache-gpac
+        uses: actions/cache@v5
+        with:
+          path: ~/gpac-install
+          key: gpac-abi-16.4-ubuntu24-deb
+
+      - name: Build GPAC
+        if: steps.should_build.outputs.should_build == 'true' && steps.cache-gpac.outputs.cache-hit != 'true'
+        run: |
+          git clone -b abi-16.4 --depth 1 https://github.com/gpac/gpac
+          cd gpac
+          ./configure --prefix=/usr
+          make -j$(nproc)
+          make DESTDIR=$HOME/gpac-install install-lib
+
+      - name: Install GPAC to system
+        if: steps.should_build.outputs.should_build == 'true'
+        run: |
+          sudo cp -r $HOME/gpac-install/usr/lib/* /usr/lib/
+          sudo cp -r $HOME/gpac-install/usr/include/* /usr/include/
+          sudo ldconfig
+
+      - name: Build CCExtractor
+        if: steps.should_build.outputs.should_build == 'true'
+        run: |
+          mkdir build && cd build
+          if [ "${{ matrix.build_type }}" = "hardsubx" ]; then
+            cmake ../src -DCMAKE_BUILD_TYPE=Release -DWITH_OCR=ON -DWITH_HARDSUBX=ON
+          else
+            cmake ../src -DCMAKE_BUILD_TYPE=Release -DWITH_OCR=ON
+          fi
+          make -j$(nproc)
+
+      - name: Test build
+        if: steps.should_build.outputs.should_build == 'true'
+        run: ./build/ccextractor --version
+
+      - name: Create .deb package structure
+        if: steps.should_build.outputs.should_build == 'true'
+        run: |
+          VERSION="${{ steps.version.outputs.version }}"
+          VARIANT="${{ matrix.build_type }}"
+
+          if [ "$VARIANT" = "basic" ]; then
+            PKG_NAME="ccextractor_${VERSION}_amd64"
+          else
+            PKG_NAME="ccextractor-${VARIANT}_${VERSION}_amd64"
+          fi
+
+          mkdir -p ${PKG_NAME}/DEBIAN
+          mkdir -p ${PKG_NAME}/usr/bin
+          mkdir -p ${PKG_NAME}/usr/lib/ccextractor
+          mkdir -p ${PKG_NAME}/usr/share/doc/ccextractor
+          mkdir -p ${PKG_NAME}/usr/share/man/man1
+
+          # Copy binary
+          cp build/ccextractor ${PKG_NAME}/usr/bin/
+
+          # Copy GPAC library
+          cp $HOME/gpac-install/usr/lib/libgpac.so* ${PKG_NAME}/usr/lib/ccextractor/
+
+          # Set rpath so ccextractor finds bundled libgpac
+          patchelf --set-rpath '/usr/lib/ccextractor:$ORIGIN/../lib/ccextractor' ${PKG_NAME}/usr/bin/ccextractor
+
+          # Copy documentation
+          cp docs/CHANGES.TXT ${PKG_NAME}/usr/share/doc/ccextractor/changelog
+          cp LICENSE.txt ${PKG_NAME}/usr/share/doc/ccextractor/copyright
+          gzip -9 -n ${PKG_NAME}/usr/share/doc/ccextractor/changelog
+
+          # Generate man page
+          help2man --no-info --name="closed captions and teletext subtitle extractor" \
+            ./build/ccextractor > ${PKG_NAME}/usr/share/man/man1/ccextractor.1 2>/dev/null || true
+          if [ -f ${PKG_NAME}/usr/share/man/man1/ccextractor.1 ]; then
+            gzip -9 -n ${PKG_NAME}/usr/share/man/man1/ccextractor.1
+          fi
+
+          # Create control file
+          if [ "$VARIANT" = "basic" ]; then
+            PKG_DESCRIPTION="CCExtractor - closed captions and teletext subtitle extractor"
+          else
+            PKG_DESCRIPTION="CCExtractor (with HardSubX) - closed captions and teletext subtitle extractor"
+          fi
+
+          INSTALLED_SIZE=$(du -sk ${PKG_NAME}/usr | cut -f1)
+
+          # Determine dependencies based on build variant (Ubuntu 24.04)
+          if [ "$VARIANT" = "hardsubx" ]; then
+            DEPENDS="libc6, libtesseract5, liblept5, libcurl3t64-gnutls, libavcodec60, libavformat60, libavutil58, libswscale7, libavdevice60, libswresample4, libavfilter9"
+          else
+            DEPENDS="libc6, libtesseract5, liblept5, libcurl3t64-gnutls"
+          fi
+
+          cat > ${PKG_NAME}/DEBIAN/control << CTRL
+          Package: ccextractor
+          Version: ${VERSION}
+          Section: utils
+          Priority: optional
+          Architecture: amd64
+          Installed-Size: ${INSTALLED_SIZE}
+          Depends: ${DEPENDS}
+          Maintainer: CCExtractor Development Team <carlos@ccextractor.org>
+          Homepage: https://www.ccextractor.org
+          Description: ${PKG_DESCRIPTION}
+           CCExtractor is a tool that extracts closed captions and teletext subtitles
+           from video files and streams. It supports a wide variety of input formats
+           including MPEG, H.264/AVC, H.265/HEVC, MP4, MKV, WTV, and transport streams.
+           .
+           This package includes a bundled GPAC library for MP4 support.
+          CTRL
+
+          # Remove leading spaces from control file
+          sed -i 's/^          //' ${PKG_NAME}/DEBIAN/control
+
+          # Create postinst to update library cache
+          cat > ${PKG_NAME}/DEBIAN/postinst << 'POSTINST'
+          #!/bin/sh
+          set -e
+          ldconfig
+          POSTINST
+          chmod 755 ${PKG_NAME}/DEBIAN/postinst
+
+          # Create postrm to update library cache
+          cat > ${PKG_NAME}/DEBIAN/postrm << 'POSTRM'
+          #!/bin/sh
+          set -e
+          ldconfig
+          POSTRM
+          chmod 755 ${PKG_NAME}/DEBIAN/postrm
+
+          # Set permissions
+          chmod 755 ${PKG_NAME}/usr/bin/ccextractor
+          chmod 755 ${PKG_NAME}/usr/lib/ccextractor
+          find ${PKG_NAME}/usr/lib/ccextractor -name "*.so*" -exec chmod 644 {} \;
+
+          # Build the .deb
+          dpkg-deb --build --root-owner-group ${PKG_NAME}
+
+          echo "deb_name=${PKG_NAME}.deb" >> $GITHUB_OUTPUT
+
+      - name: Test .deb package
+        if: steps.should_build.outputs.should_build == 'true'
+        run: |
+          VERSION="${{ steps.version.outputs.version }}"
+          VARIANT="${{ matrix.build_type }}"
+
+          if [ "$VARIANT" = "basic" ]; then
+            PKG_NAME="ccextractor_${VERSION}_amd64"
+          else
+            PKG_NAME="ccextractor-${VARIANT}_${VERSION}_amd64"
+          fi
+
+          # Install and test (apt handles dependencies automatically)
+          sudo apt-get update
+          sudo apt-get install -y ./${PKG_NAME}.deb
+          ccextractor --version
+
+      - name: Get .deb filename
+        if: steps.should_build.outputs.should_build == 'true'
+        id: deb_name
+        run: |
+          VERSION="${{ steps.version.outputs.version }}"
+          VARIANT="${{ matrix.build_type }}"
+
+          if [ "$VARIANT" = "basic" ]; then
+            echo "name=ccextractor_${VERSION}_amd64.deb" >> $GITHUB_OUTPUT
+          else
+            echo "name=ccextractor-${VARIANT}_${VERSION}_amd64.deb" >> $GITHUB_OUTPUT
+          fi
+
+      - name: Upload .deb artifact
+        if: steps.should_build.outputs.should_build == 'true'
+        uses: actions/upload-artifact@v6
+        with:
+          name: ${{ steps.deb_name.outputs.name }}
+          path: ${{ steps.deb_name.outputs.name }}
+
+      - name: Upload to Release
+        if: steps.should_build.outputs.should_build == 'true' && github.event_name == 'release'
+        uses: softprops/action-gh-release@v2
+        with:
+          files: ${{ steps.deb_name.outputs.name }}
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/build_deb_debian13.yml

@@ -0,0 +1,275 @@
+name: Build Debian 13 .deb Package
+
+on:
+  # Build on releases
+  release:
+    types: [published]
+  # Allow manual trigger
+  workflow_dispatch:
+    inputs:
+      build_type:
+        description: 'Build type (all, basic, hardsubx)'
+        required: false
+        default: 'all'
+  # Build on pushes to workflow file for testing
+  push:
+    paths:
+      - '.github/workflows/build_deb_debian13.yml'
+
+jobs:
+  build-deb:
+    runs-on: ubuntu-latest
+    container:
+      image: debian:trixie
+    strategy:
+      fail-fast: false
+      matrix:
+        build_type: [basic, hardsubx]
+
+    steps:
+      - name: Check if should build this variant
+        id: should_build
+        run: |
+          if [ "${{ github.event_name }}" = "workflow_dispatch" ]; then
+            INPUT_TYPE="${{ github.event.inputs.build_type }}"
+            if [ "$INPUT_TYPE" = "all" ] || [ "$INPUT_TYPE" = "${{ matrix.build_type }}" ]; then
+              echo "should_build=true" >> $GITHUB_OUTPUT
+            else
+              echo "should_build=false" >> $GITHUB_OUTPUT
+            fi
+          else
+            echo "should_build=true" >> $GITHUB_OUTPUT
+          fi
+
+      - name: Install git and dependencies for checkout
+        if: steps.should_build.outputs.should_build == 'true'
+        run: |
+          apt-get update
+          apt-get install -y git ca-certificates
+
+      - name: Checkout repository
+        if: steps.should_build.outputs.should_build == 'true'
+        uses: actions/checkout@v6
+
+      - name: Get version
+        if: steps.should_build.outputs.should_build == 'true'
+        id: version
+        run: |
+          # Extract version from source or use tag
+          if [ "${{ github.event_name }}" = "release" ]; then
+            VERSION="${{ github.event.release.tag_name }}"
+            VERSION="${VERSION#v}"  # Remove 'v' prefix if present
+          else
+            # Extract version from lib_ccx.h (e.g., #define VERSION "0.96.5")
+            VERSION=$(grep -oP '#define VERSION "\K[^"]+' src/lib_ccx/lib_ccx.h || echo "0.96")
+          fi
+          echo "version=$VERSION" >> $GITHUB_OUTPUT
+          echo "Building version: $VERSION"
+
+      - name: Install base dependencies
+        if: steps.should_build.outputs.should_build == 'true'
+        run: |
+          apt-get install -y --no-install-recommends \
+            build-essential \
+            cmake \
+            pkg-config \
+            zlib1g-dev \
+            libpng-dev \
+            libjpeg-dev \
+            libfreetype-dev \
+            libxml2-dev \
+            libcurl4-gnutls-dev \
+            libssl-dev \
+            clang \
+            libclang-dev \
+            tesseract-ocr \
+            libtesseract-dev \
+            libleptonica-dev \
+            patchelf \
+            curl
+
+      - name: Install FFmpeg dependencies (HardSubX)
+        if: steps.should_build.outputs.should_build == 'true' && matrix.build_type == 'hardsubx'
+        run: |
+          apt-get install -y --no-install-recommends \
+            libavcodec-dev \
+            libavformat-dev \
+            libavutil-dev \
+            libswscale-dev \
+            libswresample-dev \
+            libavfilter-dev \
+            libavdevice-dev
+
+      - name: Install Rust toolchain
+        if: steps.should_build.outputs.should_build == 'true'
+        run: |
+          curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y
+          echo "$HOME/.cargo/bin" >> $GITHUB_PATH
+
+      - name: Build GPAC
+        if: steps.should_build.outputs.should_build == 'true'
+        run: |
+          git clone -b abi-16.4 --depth 1 https://github.com/gpac/gpac
+          cd gpac
+          ./configure --prefix=/usr
+          make -j$(nproc)
+          make install-lib
+          ldconfig
+
+      - name: Build CCExtractor
+        if: steps.should_build.outputs.should_build == 'true'
+        run: |
+          export PATH="$HOME/.cargo/bin:$PATH"
+          mkdir build && cd build
+          if [ "${{ matrix.build_type }}" = "hardsubx" ]; then
+            cmake ../src -DCMAKE_BUILD_TYPE=Release -DWITH_OCR=ON -DWITH_HARDSUBX=ON
+          else
+            cmake ../src -DCMAKE_BUILD_TYPE=Release -DWITH_OCR=ON
+          fi
+          make -j$(nproc)
+
+      - name: Test build
+        if: steps.should_build.outputs.should_build == 'true'
+        run: ./build/ccextractor --version
+
+      - name: Create .deb package structure
+        if: steps.should_build.outputs.should_build == 'true'
+        id: create_deb
+        run: |
+          VERSION="${{ steps.version.outputs.version }}"
+          VARIANT="${{ matrix.build_type }}"
+
+          if [ "$VARIANT" = "basic" ]; then
+            PKG_NAME="ccextractor_${VERSION}_debian13_amd64"
+          else
+            PKG_NAME="ccextractor-${VARIANT}_${VERSION}_debian13_amd64"
+          fi
+
+          mkdir -p ${PKG_NAME}/DEBIAN
+          mkdir -p ${PKG_NAME}/usr/bin
+          mkdir -p ${PKG_NAME}/usr/lib/ccextractor
+          mkdir -p ${PKG_NAME}/usr/share/doc/ccextractor
+          mkdir -p ${PKG_NAME}/usr/share/man/man1
+
+          # Copy binary
+          cp build/ccextractor ${PKG_NAME}/usr/bin/
+
+          # Copy GPAC library
+          cp /usr/lib/libgpac.so* ${PKG_NAME}/usr/lib/ccextractor/
+
+          # Set rpath so ccextractor finds bundled libgpac
+          patchelf --set-rpath '/usr/lib/ccextractor:$ORIGIN/../lib/ccextractor' ${PKG_NAME}/usr/bin/ccextractor
+
+          # Copy documentation
+          cp docs/CHANGES.TXT ${PKG_NAME}/usr/share/doc/ccextractor/changelog
+          cp LICENSE.txt ${PKG_NAME}/usr/share/doc/ccextractor/copyright
+          gzip -9 -n ${PKG_NAME}/usr/share/doc/ccextractor/changelog
+
+          # Create control file
+          if [ "$VARIANT" = "basic" ]; then
+            PKG_DESCRIPTION="CCExtractor - closed captions and teletext subtitle extractor"
+          else
+            PKG_DESCRIPTION="CCExtractor (with HardSubX) - closed captions and teletext subtitle extractor"
+          fi
+
+          INSTALLED_SIZE=$(du -sk ${PKG_NAME}/usr | cut -f1)
+
+          # Determine dependencies based on build variant (Debian 13 Trixie)
+          if [ "$VARIANT" = "hardsubx" ]; then
+            DEPENDS="libc6, libtesseract5, libleptonica6, libcurl3t64-gnutls, libavcodec61, libavformat61, libavutil59, libswscale8, libavdevice61, libswresample5, libavfilter10"
+          else
+            DEPENDS="libc6, libtesseract5, libleptonica6, libcurl3t64-gnutls"
+          fi
+
+          cat > ${PKG_NAME}/DEBIAN/control << CTRL
+          Package: ccextractor
+          Version: ${VERSION}
+          Section: utils
+          Priority: optional
+          Architecture: amd64
+          Installed-Size: ${INSTALLED_SIZE}
+          Depends: ${DEPENDS}
+          Maintainer: CCExtractor Development Team <carlos@ccextractor.org>
+          Homepage: https://www.ccextractor.org
+          Description: ${PKG_DESCRIPTION}
+           CCExtractor is a tool that extracts closed captions and teletext subtitles
+           from video files and streams. It supports a wide variety of input formats
+           including MPEG, H.264/AVC, H.265/HEVC, MP4, MKV, WTV, and transport streams.
+           .
+           This package includes a bundled GPAC library for MP4 support.
+           Built for Debian 13 (Trixie).
+          CTRL
+
+          # Remove leading spaces from control file
+          sed -i 's/^          //' ${PKG_NAME}/DEBIAN/control
+
+          # Create postinst to update library cache
+          cat > ${PKG_NAME}/DEBIAN/postinst << 'POSTINST'
+          #!/bin/sh
+          set -e
+          ldconfig
+          POSTINST
+          chmod 755 ${PKG_NAME}/DEBIAN/postinst
+
+          # Create postrm to update library cache
+          cat > ${PKG_NAME}/DEBIAN/postrm << 'POSTRM'
+          #!/bin/sh
+          set -e
+          ldconfig
+          POSTRM
+          chmod 755 ${PKG_NAME}/DEBIAN/postrm
+
+          # Set permissions
+          chmod 755 ${PKG_NAME}/usr/bin/ccextractor
+          chmod 755 ${PKG_NAME}/usr/lib/ccextractor
+          find ${PKG_NAME}/usr/lib/ccextractor -name "*.so*" -exec chmod 644 {} \;
+
+          # Build the .deb
+          dpkg-deb --build --root-owner-group ${PKG_NAME}
+
+          echo "deb_name=${PKG_NAME}.deb" >> $GITHUB_OUTPUT
+
+      - name: Test .deb package
+        if: steps.should_build.outputs.should_build == 'true'
+        run: |
+          VERSION="${{ steps.version.outputs.version }}"
+          VARIANT="${{ matrix.build_type }}"
+
+          if [ "$VARIANT" = "basic" ]; then
+            PKG_NAME="ccextractor_${VERSION}_debian13_amd64"
+          else
+            PKG_NAME="ccextractor-${VARIANT}_${VERSION}_debian13_amd64"
+          fi
+
+          # Install and test (apt handles dependencies automatically)
+          apt-get update
+          apt-get install -y ./${PKG_NAME}.deb
+          ccextractor --version
+
+      - name: Get .deb filename
+        if: steps.should_build.outputs.should_build == 'true'
+        id: deb_name
+        run: |
+          VERSION="${{ steps.version.outputs.version }}"
+          VARIANT="${{ matrix.build_type }}"
+
+          if [ "$VARIANT" = "basic" ]; then
+            echo "name=ccextractor_${VERSION}_debian13_amd64.deb" >> $GITHUB_OUTPUT
+          else
+            echo "name=ccextractor-${VARIANT}_${VERSION}_debian13_amd64.deb" >> $GITHUB_OUTPUT
+          fi
+
+      - name: Upload .deb artifact
+        if: steps.should_build.outputs.should_build == 'true'
+        uses: actions/upload-artifact@v6
+        with:
+          name: ${{ steps.deb_name.outputs.name }}
+          path: ${{ steps.deb_name.outputs.name }}
+
+      - name: Upload to Release
+        if: steps.should_build.outputs.should_build == 'true' && github.event_name == 'release'
+        uses: softprops/action-gh-release@v2
+        with:
+          files: ${{ steps.deb_name.outputs.name }}
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/build_docker.yml

@@ -8,6 +8,8 @@ on:
     - 'docker/**'
     - '**.c'
     - '**.h'
+    - '**CMakeLists.txt'
+    - '**.cmake'
     - 'src/rust/**'
   pull_request:
     types: [opened, synchronize, reopened]
@@ -16,6 +18,8 @@ on:
     - 'docker/**'
     - '**.c'
     - '**.h'
+    - '**CMakeLists.txt'
+    - '**.cmake'
     - 'src/rust/**'
 
 jobs:

.github/workflows/build_linux.yml

@@ -7,6 +7,8 @@ on:
     - '.github/workflows/build_linux.yml'
     - '**.c'
     - '**.h'
+    - '**CMakeLists.txt'
+    - '**.cmake'
     - '**Makefile**'
     - 'linux/**'
     - 'package_creators/**'
@@ -17,6 +19,8 @@ on:
     - '.github/workflows/build_linux.yml'
     - '**.c'
     - '**.h'
+    - '**CMakeLists.txt'
+    - '**.cmake'
     - '**Makefile**'
     - 'linux/**'
     - 'package_creators/**'

.github/workflows/build_mac.yml

@@ -7,6 +7,8 @@ on:
     - '.github/workflows/build_mac.yml'
     - '**.c'
     - '**.h'
+    - '**CMakeLists.txt'
+    - '**.cmake'
     - '**Makefile**'
     - 'mac/**'
     - 'package_creators/**'
@@ -17,6 +19,8 @@ on:
     - '.github/workflows/build_mac.yml'
     - '**.c'
     - '**.h'
+    - '**CMakeLists.txt'
+    - '**.cmake'
     - '**Makefile**'
     - 'mac/**'
     - 'package_creators/**'

.github/workflows/build_snap.yml

@@ -0,0 +1,51 @@
+name: Build CCExtractor Snap
+
+on:
+  workflow_dispatch:
+  release:
+    types: [published]
+
+jobs:
+  build_snap:
+    name: Build Snap package
+    runs-on: ubuntu-22.04
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v6
+
+      - name: Install snapd
+        run: |
+          sudo apt update
+          sudo apt install -y snapd
+
+      - name: Start snapd
+        run: |
+          sudo systemctl start snapd.socket
+          sudo systemctl start snapd
+
+      - name: Install Snapcraft
+        run: |
+          sudo snap install core22
+          sudo snap install snapcraft --classic
+
+      - name: Show Snapcraft version
+        run: snapcraft --version
+
+      - name: Build snap
+        run: sudo snapcraft --destructive-mode
+
+      - name: List generated snap
+        run: ls -lh *.snap
+
+      - name: Upload snap as workflow artifact
+        uses: actions/upload-artifact@v6
+        with:
+          name: CCExtractor Snap
+          path: "*.snap"
+
+      - name: Upload snap to GitHub Release
+        if: github.event_name == 'release'
+        uses: softprops/action-gh-release@v2
+        with:
+          files: "*.snap"

.github/workflows/build_windows.yml

@@ -3,7 +3,6 @@ name: Build CCExtractor on Windows
 env:
   RUSTFLAGS: -Ctarget-feature=+crt-static
   VCPKG_DEFAULT_TRIPLET: x64-windows-static
-  VCPKG_DEFAULT_BINARY_CACHE: C:\vcpkg\.cache
   VCPKG_COMMIT: ab2977be50c702126336e5088f4836060733c899
 
 on:
@@ -13,6 +12,8 @@ on:
       - ".github/workflows/build_windows.yml"
       - "**.c"
       - "**.h"
+      - "**CMakeLists.txt"
+      - "**.cmake"
       - "windows/**"
       - "src/rust/**"
   pull_request:
@@ -21,108 +22,118 @@ on:
       - ".github/workflows/build_windows.yml"
       - "**.c"
       - "**.h"
+      - "**CMakeLists.txt"
+      - "**.cmake"
       - "windows/**"
       - "src/rust/**"
 
 jobs:
-  build_release:
+  build:
     runs-on: windows-2022
     steps:
       - name: Check out repository
         uses: actions/checkout@v6
+
       - name: Setup MSBuild.exe
         uses: microsoft/setup-msbuild@v2.0.0
         with:
           msbuild-architecture: x64
+
+      # Install GPAC (fast, ~30s, not worth caching complexity)
       - name: Install gpac
-        run: choco install gpac --version 2.4.0
+        run: choco install gpac --version 2.4.0 --no-progress
+
+      # Use lukka/run-vcpkg for better caching
       - name: Setup vcpkg
-        run: mkdir C:\vcpkg\.cache
-      - name: Cache vcpkg
-        id: cache
+        uses: lukka/run-vcpkg@v11
+        id: runvcpkg
+        with:
+          vcpkgGitCommitId: ${{ env.VCPKG_COMMIT }}
+          vcpkgDirectory: ${{ github.workspace }}/vcpkg
+          vcpkgJsonGlob: 'windows/vcpkg.json'
+
+      # Cache vcpkg installed packages separately for faster restores
+      - name: Cache vcpkg installed packages
+        id: vcpkg-installed-cache
+        uses: actions/cache@v5
+        with:
+          path: ${{ github.workspace }}/vcpkg/installed
+          key: vcpkg-installed-${{ runner.os }}-${{ env.VCPKG_COMMIT }}-${{ hashFiles('windows/vcpkg.json') }}
+          restore-keys: |
+            vcpkg-installed-${{ runner.os }}-${{ env.VCPKG_COMMIT }}-
+
+      - name: Install vcpkg dependencies
+        if: steps.vcpkg-installed-cache.outputs.cache-hit != 'true'
+        run: ${{ github.workspace }}/vcpkg/vcpkg.exe install --x-install-root ${{ github.workspace }}/vcpkg/installed/
+        working-directory: windows
+
+      # Cache Rust/Cargo artifacts
+      - name: Cache Cargo registry
         uses: actions/cache@v5
         with:
           path: |
-            C:\vcpkg\.cache
-          key: vcpkg-${{ runner.os }}-${{ env.VCPKG_COMMIT }}
-      - name: Build vcpkg
-        run: |
-          git clone https://github.com/microsoft/vcpkg
-          ./vcpkg/bootstrap-vcpkg.bat
-      - name: Install dependencies
-        run: ${{ github.workspace }}/vcpkg/vcpkg.exe install --x-install-root ${{ github.workspace }}/vcpkg/installed/
-        working-directory: windows
-      - uses: actions-rs/toolchain@v1
+            ~/.cargo/registry
+            ~/.cargo/git
+          key: ${{ runner.os }}-cargo-registry-${{ hashFiles('**/Cargo.lock') }}
+          restore-keys: |
+            ${{ runner.os }}-cargo-registry-
+
+      # Cache Cargo build artifacts - rust.bat sets CARGO_TARGET_DIR to windows/
+      # which results in artifacts at windows/x86_64-pc-windows-msvc/
+      - name: Cache Cargo build artifacts
+        uses: actions/cache@v5
         with:
-          toolchain: stable
-          override: true
+          path: ${{ github.workspace }}/windows/x86_64-pc-windows-msvc
+          key: ${{ runner.os }}-cargo-build-${{ hashFiles('**/Cargo.lock') }}-${{ hashFiles('src/rust/**/*.rs') }}
+          restore-keys: |
+            ${{ runner.os }}-cargo-build-${{ hashFiles('**/Cargo.lock') }}-
+            ${{ runner.os }}-cargo-build-
+
+      - name: Setup Rust toolchain
+        uses: dtolnay/rust-toolchain@stable
+
       - name: Install Win 10 SDK
         uses: ilammy/msvc-dev-cmd@v1
-      - name: build Release-Full
+
+      # Build Release-Full
+      - name: Build Release-Full
         env:
           LIBCLANG_PATH: "C:\\Program Files\\LLVM\\lib"
           LLVM_CONFIG_PATH: "C:\\Program Files\\LLVM\\bin\\llvm-config"
-          CARGO_TARGET_DIR: "..\\..\\windows"
           BINDGEN_EXTRA_CLANG_ARGS: -fmsc-version=0
           VCPKG_ROOT: ${{ github.workspace }}/vcpkg
         run: msbuild ccextractor.sln /p:Configuration=Release-Full /p:Platform=x64
         working-directory: ./windows
-      - name: Display version information
+
+      - name: Display Release version information
         run: ./ccextractorwinfull.exe --version
         working-directory: ./windows/x64/Release-Full
-      - uses: actions/upload-artifact@v6
+
+      - name: Upload Release artifact
+        uses: actions/upload-artifact@v6
         with:
           name: CCExtractor Windows Release build
           path: |
             ./windows/x64/Release-Full/ccextractorwinfull.exe
             ./windows/x64/Release-Full/*.dll
-  build_debug:
-    runs-on: windows-2022
-    steps:
-      - name: Check out repository
-        uses: actions/checkout@v6
-      - name: Setup MSBuild.exe
-        uses: microsoft/setup-msbuild@v2.0.0
-        with:
-          msbuild-architecture: x64
-      - name: Install gpac
-        run: choco install gpac --version 2.4.0
-      - name: Setup vcpkg
-        run: mkdir C:\vcpkg\.cache
-      - name: Cache vcpkg
-        id: cache
-        uses: actions/cache@v5
-        with:
-          path: |
-            C:\vcpkg\.cache
-          key: vcpkg-${{ runner.os }}-${{ env.VCPKG_COMMIT }}
-      - name: Build vcpkg
-        run: |
-          git clone https://github.com/microsoft/vcpkg
-          ./vcpkg/bootstrap-vcpkg.bat
-      - name: Install dependencies
-        run: ${{ github.workspace }}/vcpkg/vcpkg.exe install --x-install-root ${{ github.workspace }}/vcpkg/installed/
-        working-directory: windows
-      - uses: actions-rs/toolchain@v1
-        with:
-          toolchain: stable
-          override: true
-      - name: Install Win 10 SDK
-        uses: ilammy/msvc-dev-cmd@v1
-      - name: build Debug-Full
+
+      # Build Debug-Full (reuses cached Cargo artifacts)
+      - name: Build Debug-Full
         env:
           LIBCLANG_PATH: "C:\\Program Files\\LLVM\\lib"
           LLVM_CONFIG_PATH: "C:\\Program Files\\LLVM\\bin\\llvm-config"
-          CARGO_TARGET_DIR: "..\\..\\windows"
           BINDGEN_EXTRA_CLANG_ARGS: -fmsc-version=0
           VCPKG_ROOT: ${{ github.workspace }}/vcpkg
         run: msbuild ccextractor.sln /p:Configuration=Debug-Full /p:Platform=x64
         working-directory: ./windows
-      - name: Display version information
+
+      - name: Display Debug version information
         continue-on-error: true
         run: ./ccextractorwinfull.exe --version
         working-directory: ./windows/x64/Debug-Full
-      - uses: actions/upload-artifact@v6
+
+      - name: Upload Debug artifact
+        uses: actions/upload-artifact@v6
         with:
           name: CCExtractor Windows Debug build
           path: |

.github/workflows/homebrew.yml

@@ -0,0 +1,15 @@
+name: Bump Homebrew Formula
+
+on:
+  release:
+    types: [published]
+
+jobs:
+  homebrew:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Update Homebrew formula
+        uses: dawidd6/action-homebrew-bump-formula@v7
+        with:
+          token: ${{ secrets.HOMEBREW_GITHUB_API_TOKEN }}
+          formula: ccextractor

.github/workflows/release.yml

@@ -109,7 +109,7 @@ jobs:
       run: Compress-Archive -Path ./installer/* -DestinationPath ./CCExtractor.${{ steps.get_version.outputs.DISPLAY_VERSION }}_win_portable.zip
       working-directory: ./windows
     - name: Build installer
-      run:  wix build -ext WixToolset.UI.wixext -d "AppVersion=${{ steps.get_version.outputs.VERSION }}" -o CCExtractor.${{ steps.get_version.outputs.DISPLAY_VERSION }}.msi installer.wxs CustomUI.wxs
+      run:  wix build -arch x64 -ext WixToolset.UI.wixext -d "AppVersion=${{ steps.get_version.outputs.VERSION }}" -o CCExtractor.${{ steps.get_version.outputs.DISPLAY_VERSION }}.msi installer.wxs CustomUI.wxs
       working-directory: ./windows
     - name: Upload as asset
       uses: AButler/upload-release-assets@v3.0

.gitignore

@@ -17,6 +17,7 @@ CVS
 mac/ccextractor
 linux/ccextractor
 linux/depend
+linux/build_scan/
 windows/x86_64-pc-windows-msvc/**
 windows/Debug/**
 windows/Debug-OCR/**
@@ -28,6 +29,7 @@ windows/Debug-Full/**
 windows/x64/**
 windows/ccextractor.VC.db
 build/
+build_*/
 
 ####
 # Python
@@ -143,6 +145,9 @@ bazel*
 #Intellij IDEs
 .idea/
 
+# Plans (local only)
+plans/
+
 # Rust build and MakeFiles (and CMake files)
 src/rust/CMakeFiles/
 src/rust/CMakeCache.txt

OpenBSD/Makefile

@@ -4,7 +4,7 @@ MAINTAINER = 	Marc Espie <espie@openbsd.org>
 CATEGORIES =	multimedia
 COMMENT =	closed caption subtitles extractor
 HOMEPAGE = 	https://ccextractor.org
-V =		0.96.3
+V =		0.96.5
 DISTFILES =	ccextractor.${V:S/.//}-src.zip
 MASTER_SITES =	${MASTER_SITE_SOURCEFORGE:=ccextractor/}
 DISTNAME =	ccextractor-$V

README.md

@@ -28,6 +28,25 @@ The core functionality is written in C. Other languages used include C++ and Pyt
 
 Downloads for precompiled binaries and source code can be found [on our website](https://ccextractor.org/public/general/downloads/).
 
+
+### Windows Package Managers
+
+**WinGet:**
+```powershell
+winget install CCExtractor.CCExtractor
+```
+
+**Chocolatey:**
+```powershell
+choco install ccextractor
+```
+
+**Scoop:**
+```powershell
+scoop bucket add extras
+scoop install ccextractor
+```
+
 Extracting subtitles is relatively simple. Just run the following command:
 
 `ccextractor <input>`
@@ -43,6 +62,34 @@ You can also find the list of parameters and their brief description by running
 
 You can find sample files on [our website](https://ccextractor.org/public/general/tvsamples/) to test the software.
 
+### Building from Source
+
+- [Building on Windows using WSL](docs/build-wsl.md)
+
+#### Linux (Autotools) build notes
+
+CCExtractor also supports an autotools-based build system under the `linux/`
+directory.
+
+Important notes:
+- The autotools workflow lives inside `linux/`. The `configure` script is
+  generated there and should be run from that directory.
+- Typical build steps are:
+```
+cd linux
+./autogen.sh
+./configure
+make
+```
+- Rust support is enabled automatically if `cargo` and `rustc` are available
+  on the system. In that case, Rust components are built and linked during
+  `make`.
+- If you encounter unexpected build or linking issues, a clean rebuild
+  (`make clean` or a fresh clone) is recommended, especially when Rust is
+  involved.
+
+This build flow has been tested on Linux and WSL.
+
 ## Compiling CCExtractor
 
 To learn more about how to compile and build CCExtractor for your platform check the [compilation guide](https://github.com/CCExtractor/ccextractor/blob/master/docs/COMPILATION.MD).

docs/CHANGES.TXT

@@ -1,3 +1,34 @@
+0.96.6 (unreleased)
+-------------------
+- New: Add Snap packaging support with Snapcraft configuration and GitHub Actions CI workflow. 
+- Fix: Clear status line output on Linux/WSL to prevent text artifacts (#2017)
+- Fix: Prevent infinite loop on truncated MKV files
+- Fix: Various memory safety and stability fixes in demuxers (MP4, PS, MKV, DVB)
+- Fix: Delete empty output files instead of leaving 0-byte files (#1282)
+- Fix: --mkvlang now supports BCP 47 language tags (e.g., en-US, zh-Hans-CN) and multiple codes
+
+0.96.5 (2026-01-05)
+-------------------
+- New: CCExtractor is available again via Homebrew on macOS and Linux.
+- New: Add support for raw CDP (Caption Distribution Packet) files (#1406)
+- New: Add --scc-accurate-timing option for bandwidth-aware SCC output (#1120)
+- Fix: MXF files containing CEA-708 captions not being detected/extracted (#1647)
+- Docs: Add Windows WSL build instructions
+- Fix: Security fixes (out-of-bounds read/write) in a few places in the legacy C code.
+
+0.96.4 (2026-01-01)
+-------------------
+- New: Persistent CEA-708 decoder context - maintains state across multiple calls for proper subtitle continuity
+- New: OCR character blacklist options (--ocr-blacklist, --ocr-blacklist-file) for improved accuracy
+- New: OCR line-split option (--ocr-splitontimechange) for better subtitle segmentation
+- Fix: 32-bit build failures on i686 and armv7l architectures
+- Fix: Legacy command-line argument compatibility (-1, -2, -12, --sc, --svc)
+- Fix: Prevent heap buffer overflow in Teletext processing (security fix)
+- Fix: Prevent integer overflow leading to heap buffer overflow in Transport Stream handling (security fix)
+- Fix: Lazy OCR initialization - only initialize when first DVB subtitle is encountered
+- Build: Optimized Windows CI workflow for faster builds
+- Fix: Updated GUI with version 0.7.1. A blind attempt to fix a hang on start on some Windows.
+
 0.96.3 (2025-12-29)
 -------------------
 - New: VOBSUB subtitle extraction with OCR support for MP4 files
@@ -31,6 +62,7 @@
   - Extract multiple teletext pages simultaneously with separate output files
   - Use --tpage multiple times (e.g., --tpage 100 --tpage 200)
   - Output files are named with page suffix (e.g., output_p100.srt, output_p200.srt)
+- Fix: SPUPNG subtitle offset calculation to center based on actual image dimensions
 
 - New: Added --list-tracks (-L) option to list all tracks in media files without processing
   New: Chinese, Korean, Japanese support - proper encoding and OCR.

docs/COMPILATION.MD

@@ -1,3 +1,16 @@
+# Installation
+
+## Homebrew 
+The easiest way to install CCExtractor for Mac and Linux is through Homebrew:
+
+```bash
+brew install ccextractor
+```
+Note: If you don't have Homebrew installed, see [brew.sh](https://brew.sh/)
+ for installation instructions.
+
+---
+
 # Compiling CCExtractor
 
 You may compile CCExtractor across all major platforms using `CMakeLists.txt` stored under `ccextractor/src/` directory. Autoconf and custom build scripts are also available. See platform specific instructions in the below sections.

docs/build-wsl.md

@@ -0,0 +1,137 @@
+# Building CCExtractor on Windows using WSL
+
+This guide explains how to build CCExtractor on Windows using WSL (Ubuntu).
+It is based on a fresh setup and includes all required dependencies and
+common build issues encountered during compilation.
+
+---
+
+## Prerequisites
+
+- Windows 10 or Windows 11
+- WSL enabled
+- Ubuntu installed via Microsoft Store
+
+---
+
+## Install WSL and Ubuntu
+
+From PowerShell (run as Administrator):
+
+```powershell
+wsl --install -d Ubuntu
+```
+
+Restart the system if prompted, then launch Ubuntu from the Start menu.
+
+---
+
+## Update system packages
+
+```bash
+sudo apt update
+```
+
+---
+
+## Install basic build tools
+
+```bash
+sudo apt install -y build-essential git pkg-config
+```
+
+---
+
+## Install Rust (required)
+
+CCExtractor includes Rust components, so Rust and Cargo are required.
+
+```bash
+curl https://sh.rustup.rs -sSf | sh
+source ~/.cargo/env
+```
+
+Verify installation:
+
+```bash
+cargo --version
+rustc --version
+```
+
+---
+
+## Install required libraries
+
+```bash
+sudo apt install -y \
+  libclang-dev clang \
+  libtesseract-dev tesseract-ocr \
+  libgpac-dev
+```
+
+---
+
+## Clone the repository
+
+```bash
+git clone https://github.com/CCExtractor/ccextractor.git
+cd ccextractor
+```
+
+---
+
+## Build CCExtractor
+
+```bash
+cd linux
+./build
+```
+
+After a successful build, verify by running:
+
+```bash
+./ccextractor
+```
+
+You should see the help/usage output.
+
+---
+
+## Common build issues
+
+### cargo: command not found
+
+```bash
+source ~/.cargo/env
+```
+
+---
+
+### Unable to find libclang
+
+```bash
+sudo apt install libclang-dev clang
+```
+
+---
+
+### gpac/isomedia.h: No such file or directory
+
+```bash
+sudo apt install libgpac-dev
+```
+
+---
+
+### please install tesseract development library
+
+```bash
+sudo apt install libtesseract-dev tesseract-ocr
+```
+
+---
+
+## Notes
+
+- Compiler warnings during the build process are expected and do not indicate failure.
+- This guide was tested on Ubuntu (WSL) running on Windows 11.

linux/configure.ac

@@ -2,7 +2,7 @@
 # Process this file with autoconf to produce a configure script.
 
 AC_PREREQ([2.71])
-AC_INIT([CCExtractor], [0.96.3], [carlos@ccextractor.org])
+AC_INIT([CCExtractor], [0.96.5], [carlos@ccextractor.org])
 AC_CONFIG_AUX_DIR([build-conf])
 AC_CONFIG_SRCDIR([../src/ccextractor.c])
 AM_INIT_AUTOMAKE([foreign subdir-objects])

mac/build.command

@@ -42,7 +42,16 @@ while [[ $# -gt 0 ]]; do
   esac
 done
 
-BLD_FLAGS="-std=gnu99 -Wno-write-strings -Wno-pointer-sign -D_FILE_OFFSET_BITS=64 -DVERSION_FILE_PRESENT -Dfopen64=fopen -Dopen64=open -Dlseek64=lseek"
+# Determine architecture based on cargo (to ensure consistency with Rust part)
+CARGO_ARCH=$(file $(which cargo) | grep -o 'x86_64\|arm64')
+if [[ "$CARGO_ARCH" == "x86_64" ]]; then
+    echo "Detected Intel (x86_64) Cargo. Forcing x86_64 build to match Rust and libraries..."
+    BLD_ARCH="-arch x86_64"
+else
+    BLD_ARCH="-arch arm64"
+fi
+
+BLD_FLAGS="$BLD_ARCH -std=gnu99 -Wno-write-strings -Wno-pointer-sign -D_FILE_OFFSET_BITS=64 -DVERSION_FILE_PRESENT -Dfopen64=fopen -Dopen64=open -Dlseek64=lseek"
 
 # Add flags for bundled libraries (not needed when using system libs)
 if [[ "$USE_SYSTEM_LIBS" != "true" ]]; then

mac/configure.ac

@@ -2,7 +2,7 @@
 # Process this file with autoconf to produce a configure script.
 
 AC_PREREQ([2.71])
-AC_INIT([CCExtractor],[0.96.3],[carlos@ccextractor.org])
+AC_INIT([CCExtractor],[0.96.5],[carlos@ccextractor.org])
 AC_CONFIG_AUX_DIR([build-conf])
 AC_CONFIG_SRCDIR([../src/ccextractor.c])
 AM_INIT_AUTOMAKE([foreign subdir-objects])

package_creators/PKGBUILD

@@ -1,5 +1,5 @@
 pkgname=ccextractor
-pkgver=0.96.3
+pkgver=0.96.5
 pkgrel=1
 pkgdesc="A closed captions and teletext subtitles extractor for video streams."
 arch=('i686' 'x86_64')

package_creators/ccextractor.spec

@@ -1,5 +1,5 @@
 Name: ccextractor
-Version: 0.96.3
+Version: 0.96.5
 Release: 1
 Summary: A closed captions and teletext subtitles extractor for video streams.
 Group: Applications/Internet

package_creators/debian.sh

@@ -1,7 +1,7 @@
 #!/bin/bash
 TYPE="debian"   # can be one of 'slackware', 'debian', 'rpm'
 PROGRAM_NAME="ccextractor"
-VERSION="0.96.3"
+VERSION="0.96.5"
 RELEASE="1"
 LICENSE="GPL-2.0"
 MAINTAINER="carlos@ccextractor.org"

packaging/chocolatey/ccextractor.nuspec

@@ -2,7 +2,7 @@
 <package xmlns="http://schemas.microsoft.com/packaging/2015/06/nuspec.xsd">
   <metadata>
     <id>ccextractor</id>
-    <version>0.96.3</version>
+    <version>0.96.5</version>
     <title>CCExtractor</title>
     <authors>CCExtractor Development Team</authors>
     <owners>CCExtractor</owners>

packaging/chocolatey/tools/chocolateyInstall.ps1

@@ -7,7 +7,7 @@ $toolsDir = "$(Split-Path -parent $MyInvocation.MyCommand.Definition)"
 $packageArgs = @{
   packageName    = $packageName
   fileType       = 'MSI'
-  url64bit       = 'https://github.com/CCExtractor/ccextractor/releases/download/v0.96.3/CCExtractor.0.96.3.msi'
+  url64bit       = 'https://github.com/CCExtractor/ccextractor/releases/download/v0.96.5/CCExtractor.0.96.5.msi'
   checksum64     = 'FFCAB0D766180AFC2832277397CDEC885D15270DECE33A9A51947B790F1F095B'
   checksumType64 = 'sha256'
   silentArgs     = '/quiet /norestart'

packaging/winget/CCExtractor.CCExtractor.installer.yaml

@@ -1,6 +1,6 @@
 # yaml-language-server: $schema=https://aka.ms/winget-manifest.installer.1.9.0.schema.json
 PackageIdentifier: CCExtractor.CCExtractor
-PackageVersion: 0.96.3
+PackageVersion: 0.96.5
 Platform:
 - Windows.Desktop
 MinimumOSVersion: 10.0.0.0
@@ -15,7 +15,7 @@ UpgradeBehavior: install
 Installers:
 - Architecture: x64
   InstallerType: msi
-  InstallerUrl: https://github.com/CCExtractor/ccextractor/releases/download/v0.96.3/CCExtractor.0.96.3.msi
+  InstallerUrl: https://github.com/CCExtractor/ccextractor/releases/download/v0.96.5/CCExtractor.0.96.5.msi
   InstallerSha256: FFCAB0D766180AFC2832277397CDEC885D15270DECE33A9A51947B790F1F095B
 ManifestType: installer
 ManifestVersion: 1.9.0

packaging/winget/CCExtractor.CCExtractor.locale.en-US.yaml

@@ -1,6 +1,6 @@
 # yaml-language-server: $schema=https://aka.ms/winget-manifest.defaultLocale.1.9.0.schema.json
 PackageIdentifier: CCExtractor.CCExtractor
-PackageVersion: 0.96.3
+PackageVersion: 0.96.5
 PackageLocale: en-US
 Publisher: CCExtractor Development
 PublisherUrl: https://ccextractor.org

packaging/winget/CCExtractor.CCExtractor.yaml

@@ -1,6 +1,6 @@
 # yaml-language-server: $schema=https://aka.ms/winget-manifest.version.1.9.0.schema.json
 PackageIdentifier: CCExtractor.CCExtractor
-PackageVersion: 0.96.3
+PackageVersion: 0.96.5
 DefaultLocale: en-US
 ManifestType: version
 ManifestVersion: 1.9.0

snap/local/run-ccextractor.sh

@@ -0,0 +1,19 @@
+#!/bin/sh
+set -e
+# Default fallback
+LIB_TRIPLET="x86_64-linux-gnu"
+# Detect multiarch directory if present
+for d in "$SNAP/usr/lib/"*-linux-gnu; do
+    if [ -d "$d" ]; then
+        LIB_TRIPLET=$(basename "$d")
+        break
+    fi
+done
+export LD_LIBRARY_PATH="$SNAP/usr/lib:\
+$SNAP/usr/lib/$LIB_TRIPLET:\
+$SNAP/usr/lib/$LIB_TRIPLET/blas:\
+$SNAP/usr/lib/$LIB_TRIPLET/lapack:\
+$SNAP/usr/lib/$LIB_TRIPLET/pulseaudio:\
+${LD_LIBRARY_PATH:+:$LD_LIBRARY_PATH}"
+shift
+exec "$SNAP/usr/local/bin/ccextractor" "$@"

snap/snapcraft.yaml

@@ -0,0 +1,104 @@
+name: ccextractor
+base: core22
+version: '0.96.5'
+summary: Closed Caption Extractor
+description: |
+  CCExtractor is a tool for extracting closed captions from video files.
+website: https://www.ccextractor.org
+source-code: https://github.com/CCExtractor/ccextractor
+confinement: classic
+
+apps:
+  ccextractor:
+    command: usr/local/bin/ccextractor
+    command-chain:
+      - local/run-ccextractor.sh
+    plugs:
+      - home
+
+parts:
+  gpac:
+    plugin: make
+    source: https://github.com/gpac/gpac.git
+    source-tag: abi-16.4
+    build-packages:
+      - build-essential
+      - pkg-config
+      - zlib1g-dev
+      - libssl-dev
+      - libfreetype6-dev
+      - libjpeg-dev
+      - libpng-dev
+    override-build: |
+      set -eux
+      ./configure --prefix=/usr
+      make -j$(nproc)
+      make DESTDIR=$SNAPCRAFT_PART_INSTALL install-lib
+      sed -i "s|^prefix=.*|prefix=$SNAPCRAFT_STAGE/usr|" $SNAPCRAFT_PART_INSTALL/usr/lib/pkgconfig/gpac.pc
+    stage:
+      - usr/lib/libgpac*
+      - usr/lib/pkgconfig/gpac.pc
+      - usr/include/gpac
+
+  ccextractor:
+    after: [gpac]
+    plugin: cmake
+    source: .
+    source-subdir: src
+    build-environment:
+      - PKG_CONFIG_PATH: "$SNAPCRAFT_STAGE/usr/lib/pkgconfig:$PKG_CONFIG_PATH"
+    build-snaps:
+      - cmake/latest/stable
+      - rustup/latest/stable
+    build-packages:
+      - build-essential
+      - pkg-config
+      - clang
+      - llvm-dev
+      - libclang-dev
+      - libzvbi-dev
+      - libtesseract-dev
+      - libavcodec-dev
+      - libavformat-dev
+      - libavdevice-dev
+      - libavfilter-dev
+      - libswscale-dev
+      - libx11-dev
+      - libxcb1-dev
+      - libxcb-shm0-dev
+      - libpng-dev
+      - zlib1g-dev
+      - libblas3
+      - liblapack3
+    stage-packages:
+      - libzvbi0
+      - libfreetype6
+      - libpng16-16
+      - libprotobuf-c1
+      - libutf8proc2
+      - libgl1
+      - libglu1-mesa
+      - libavcodec58
+      - libavformat58
+      - libavutil56
+      - libavdevice58
+      - libavfilter7
+      - libswscale5
+      - libjpeg-turbo8
+      - libvorbis0a
+      - libtheora0
+      - libxvidcore4
+      - libfaad2
+      - libmad0
+      - liba52-0.7.4
+      - libpulse0
+      - pulseaudio-utils
+    override-build: |
+      set -eux
+      rustup toolchain install stable
+      rustup default stable
+      export PATH="$HOME/.cargo/bin:$PATH"
+      snapcraftctl build
+      install -D -m 0755 \
+        $SNAPCRAFT_PROJECT_DIR/snap/local/run-ccextractor.sh \
+        $SNAPCRAFT_PART_INSTALL/local/run-ccextractor.sh

src/CMakeLists.txt

@@ -9,7 +9,7 @@ option (WITH_HARDSUBX "Build with support for burned-in subtitles" OFF)
 
 # Version number
 set (CCEXTRACTOR_VERSION_MAJOR 0)
-set (CCEXTRACTOR_VERSION_MINOR 89)
+set (CCEXTRACTOR_VERSION_MINOR 96)
 
 # Get project directory
 get_filename_component(BASE_PROJ_DIR ../ ABSOLUTE)
@@ -255,4 +255,13 @@ endif (PKG_CONFIG_FOUND)
 target_link_libraries (ccextractor ${EXTRA_LIBS})
 target_include_directories (ccextractor PUBLIC ${EXTRA_INCLUDES})
 
+# ccx_rust (Rust) calls C functions from ccx (like decode_vbi).
+# Force the linker to pull these symbols from ccx before processing ccx_rust.
+if (NOT WIN32 AND NOT APPLE)
+  target_link_options (ccextractor PRIVATE
+    -Wl,--undefined=decode_vbi
+    -Wl,--undefined=do_cb
+    -Wl,--undefined=store_hdcc)
+endif()
+
 install (TARGETS ccextractor DESTINATION bin)

src/ccextractor.c

@@ -435,6 +435,9 @@ int main(int argc, char *argv[])
 
 	int compile_ret = ccxr_parse_parameters(argc, argv);
 
+	// Update the Rust logger target after parsing so --quiet is respected
+	ccxr_update_logger_target();
+
 	if (compile_ret == EXIT_NO_INPUT_FILES)
 	{
 		print_usage();

src/lib_ccx/CMakeLists.txt

@@ -1,9 +1,9 @@
 cmake_policy (SET CMP0037 NEW)
 
 if(MSVC)
-    set (CMAKE_C_FLAGS "-W3 /wd4005 /wd4996")
+    set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -W3 /wd4005 /wd4996")
 else (MSVC)
-    set (CMAKE_C_FLAGS "-Wall -Wno-pointer-sign -g -std=gnu99")
+    set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wall -Wno-pointer-sign -g -std=gnu99")
 endif(MSVC)
 
 if(WIN32)

src/lib_ccx/avc_functions.c

@@ -379,11 +379,10 @@ void sei_rbsp(struct avc_ctx *ctx, unsigned char *seibuf, unsigned char *seiend)
 	}
 	else
 	{
-		// TODO: This really really looks bad
-		mprint("WARNING: Unexpected SEI unit length...trying to continue.");
-		temp_debug = 1;
-		mprint("\n Failed block (at sei_rbsp) was:\n");
-		dump(CCX_DMT_GENERIC_NOTICES, (unsigned char *)seibuf, seiend - seibuf, 0, 0);
+		// Unexpected SEI length - common with malformed streams, don't spam output
+		dbg_print(CCX_DMT_VERBOSE, "WARNING: Unexpected SEI unit length (parsed to %p, expected %p)...trying to continue.\n",
+			  (void *)tbuf, (void *)(seiend - 1));
+		dump(CCX_DMT_VERBOSE, (unsigned char *)seibuf, seiend - seibuf, 0, 0);
 
 		ctx->num_unexpected_sei_length++;
 	}
@@ -393,20 +392,24 @@ void sei_rbsp(struct avc_ctx *ctx, unsigned char *seibuf, unsigned char *seiend)
 unsigned char *sei_message(struct avc_ctx *ctx, unsigned char *seibuf, unsigned char *seiend)
 {
 	int payload_type = 0;
-	while (*seibuf == 0xff)
+	while (seibuf < seiend && *seibuf == 0xff)
 	{
 		payload_type += 255;
 		seibuf++;
 	}
+	if (seibuf >= seiend)
+		return NULL;
 	payload_type += *seibuf;
 	seibuf++;
 
 	int payload_size = 0;
-	while (*seibuf == 0xff)
+	while (seibuf < seiend && *seibuf == 0xff)
 	{
 		payload_size += 255;
 		seibuf++;
 	}
+	if (seibuf >= seiend)
+		return NULL;
 	payload_size += *seibuf;
 	seibuf++;
 
@@ -904,10 +907,10 @@ void seq_parameter_set_rbsp(struct avc_ctx *ctx, unsigned char *seqbuf, unsigned
 		dvprint("vcl_hrd_parameters_present_flag=               %llX\n", tmp1);
 		if (tmp)
 		{
-			// TODO.
-			mprint("vcl_hrd. Not implemented for now. Hopefully not needed. Skipping rest of NAL\n");
+			// VCL HRD parameters are for video buffering compliance, not needed for caption extraction.
+			// Just skip and continue - this doesn't affect our ability to extract captions.
+			mprint("Skipping VCL HRD parameters (not needed for caption extraction)\n");
 			ctx->num_vcl_hrd++;
-			// exit(1);
 		}
 		if (tmp || tmp1)
 		{
@@ -993,9 +996,9 @@ void slice_header(struct encoder_ctx *enc_ctx, struct lib_cc_decode *dec_ctx, un
 
 	if (nal_unit_type == 5)
 	{
+		// idr_pic_id: Read to advance bitstream position; value not needed for caption extraction
 		tmp = read_exp_golomb_unsigned(&q1);
 		dvprint("idr_pic_id=            % 4lld (%#llX)\n", tmp, tmp);
-		// TODO
 	}
 	if (dec_ctx->avc_ctx->pic_order_cnt_type == 0)
 	{

src/lib_ccx/ccx_common_option.c

@@ -74,6 +74,8 @@ void init_options(struct ccx_s_options *options)
 	options->ocr_oem = -1;		  // By default, OEM mode depends on the tesseract version
 	options->psm = 3;		  // Default PSM mode (3 is the default tesseract as well)
 	options->ocr_quantmode = 0;	  // No quantization (better OCR accuracy for DVB subtitles)
+	options->ocr_line_split = 0;	  // By default, don't split images into lines (pending testing)
+	options->ocr_blacklist = 1;	  // By default, use character blacklist to prevent common OCR errors (| vs I, etc.)
 	options->mkvlang = NULL;	  // By default, all the languages are extracted
 	options->ignore_pts_jumps = 1;
 	options->analyze_video_stream = 0;
@@ -139,8 +141,9 @@ void init_options(struct ccx_s_options *options)
 	options->enc_cfg.services_charsets = NULL;
 	options->enc_cfg.all_services_charset = NULL;
 	options->enc_cfg.with_semaphore = 0;
-	options->enc_cfg.force_dropframe = 0; // Assume No Drop Frame for MCC Encode.
-	options->enc_cfg.scc_framerate = 0;   // Default: 29.97fps for SCC output
+	options->enc_cfg.force_dropframe = 0;	  // Assume No Drop Frame for MCC Encode.
+	options->enc_cfg.scc_framerate = 0;	  // Default: 29.97fps for SCC output
+	options->enc_cfg.scc_accurate_timing = 0; // Default: off for backwards compatibility (issue #1120)
 	options->enc_cfg.extract_only_708 = 0;
 
 	options->settings_dtvcc.enabled = 0;

src/lib_ccx/ccx_common_option.h

@@ -76,7 +76,8 @@ struct encoder_cfg
 	int force_dropframe; // 1 if dropframe frame count should be used. defaults to no drop frame.
 
 	// SCC output framerate
-	int scc_framerate; // SCC output framerate: 0=29.97 (default), 1=24, 2=25, 3=30
+	int scc_framerate;	 // SCC output framerate: 0=29.97 (default), 1=24, 2=25, 3=30
+	int scc_accurate_timing; // If 1, use bandwidth-aware timing for broadcast compliance (issue #1120)
 
 	// text -> png (text render)
 	char *render_font; // The font used to render text if needed (e.g. teletext->spupng)
@@ -152,6 +153,8 @@ struct ccx_s_options // Options from user parameters
 	int ocr_oem;		  // The Tesseract OEM mode, could be 0 (default), 1 or 2
 	int psm;		  // The Tesseract PSM mode, could be between 0 and 13. 3 is tesseract default
 	int ocr_quantmode;	  // How to quantize the bitmap before passing to to tesseract (0=no quantization at all, 1=CCExtractor's internal)
+	int ocr_line_split;	  // If 1, split images into lines before OCR (uses PSM 7 for better accuracy)
+	int ocr_blacklist;	  // If 1, use character blacklist to prevent common OCR errors (default: enabled)
 	char *mkvlang;		  // The name of the language stream for MKV
 	int analyze_video_stream; // If 1, the video stream will be processed even if we're using a different one for subtitles.
 

src/lib_ccx/ccx_decoders_608.c

@@ -201,6 +201,9 @@ void delete_to_end_of_row(ccx_decoder_608_context *context)
 {
 	if (context->mode != MODE_TEXT)
 	{
+		if (context->cursor_row >= CCX_DECODER_608_SCREEN_ROWS)
+			return;
+
 		struct eia608_screen *use_buffer = get_writing_buffer(context);
 		for (int i = context->cursor_column; i <= CCX_DECODER_608_SCREEN_WIDTH - 1; i++)
 		{
@@ -221,6 +224,10 @@ void write_char(const unsigned char c, ccx_decoder_608_context *context)
 		/* printf ("\rWriting char [%c] at %s:%d:%d\n",c,
 		use_buffer == &wb->data608->buffer1?"B1":"B2",
 		wb->data608->cursor_row,wb->data608->cursor_column); */
+
+		if (context->cursor_row >= CCX_DECODER_608_SCREEN_ROWS || context->cursor_column >= CCX_DECODER_608_SCREEN_WIDTH)
+			return;
+
 		use_buffer->characters[context->cursor_row][context->cursor_column] = c;
 		use_buffer->colors[context->cursor_row][context->cursor_column] = context->current_color;
 		use_buffer->fonts[context->cursor_row][context->cursor_column] = context->font;
@@ -316,10 +323,20 @@ int write_cc_buffer(ccx_decoder_608_context *context, struct cc_subtitle *sub)
 
 	if (!data->empty && context->output_format != CCX_OF_NULL)
 	{
-		struct eia608_screen *new_data = (struct eia608_screen *)realloc(sub->data, (sub->nb_data + 1) * sizeof(*data));
+		size_t new_size;
+
+		if (sub->nb_data + 1 > SIZE_MAX / sizeof(struct eia608_screen))
+		{
+			ccx_common_logging.log_ftn("Too many screens, cannot allocate more memory.\n");
+			return 0;
+		}
+
+		new_size = (sub->nb_data + 1) * sizeof(struct eia608_screen);
+
+		struct eia608_screen *new_data = (struct eia608_screen *)realloc(sub->data, new_size);
 		if (!new_data)
 		{
-			ccx_common_logging.log_ftn("No Memory left");
+			ccx_common_logging.log_ftn("Out of memory while reallocating screen buffer\n");
 			return 0;
 		}
 		sub->data = new_data;
@@ -386,10 +403,20 @@ int write_cc_line(ccx_decoder_608_context *context, struct cc_subtitle *sub)
 
 	if (!data->empty)
 	{
-		struct eia608_screen *new_data = (struct eia608_screen *)realloc(sub->data, (sub->nb_data + 1) * sizeof(*data));
+		size_t new_size;
+
+		if (sub->nb_data + 1 > SIZE_MAX / sizeof(struct eia608_screen))
+		{
+			ccx_common_logging.log_ftn("Too many screens, cannot allocate more memory.\n");
+			return 0;
+		}
+
+		new_size = (sub->nb_data + 1) * sizeof(struct eia608_screen);
+
+		struct eia608_screen *new_data = (struct eia608_screen *)realloc(sub->data, new_size);
 		if (!new_data)
 		{
-			ccx_common_logging.log_ftn("No Memory left");
+			ccx_common_logging.log_ftn("Out of memory while reallocating screen buffer\n");
 			return 0;
 		}
 		sub->data = new_data;

src/lib_ccx/ccx_decoders_708.c

@@ -998,6 +998,14 @@ void dtvcc_handle_DFx_DefineWindow(dtvcc_service_decoder *decoder, int window_id
 	int row_count = (data[4] & 0xf) + 1; // according to CEA-708-D
 	int anchor_point = data[4] >> 4;
 	int col_count = (data[5] & 0x3f) + 1; // according to CEA-708-D
+
+	if (row_count > CCX_DTVCC_MAX_ROWS || col_count > CCX_DTVCC_MAX_COLUMNS)
+	{
+		ccx_common_logging.log_ftn("[CEA-708] Invalid window size %dx%d (max %dx%d), rejecting window definition\n",
+					   row_count, col_count, CCX_DTVCC_MAX_ROWS, CCX_DTVCC_MAX_COLUMNS);
+		return;
+	}
+
 	int pen_style = data[6] & 0x7;
 	int win_style = (data[6] >> 3) & 0x7;
 
@@ -1341,6 +1349,14 @@ void dtvcc_handle_SPL_SetPenLocation(dtvcc_service_decoder *decoder, unsigned ch
 	}
 
 	dtvcc_window *window = &decoder->windows[decoder->current_window];
+	if (row >= window->row_count || col >= window->col_count)
+	{
+		ccx_common_logging.log_ftn("[CEA-708] dtvcc_handle_SPL_SetPenLocation: "
+					   "Invalid pen location %d:%d for window size %dx%d, rejecting command\n",
+					   row, col, window->row_count, window->col_count);
+		return;
+	}
+
 	window->pen_row = row;
 	window->pen_column = col;
 }
@@ -1479,7 +1495,12 @@ int dtvcc_handle_C0(dtvcc_ctx *dtvcc,
 	else if (c0 >= 0x18 && c0 <= 0x1F)
 	{
 		if (c0 == DTVCC_C0_P16) // PE16
-			dtvcc_handle_C0_P16(decoder, data + 1);
+		{
+			if (data_length >= 3)
+				dtvcc_handle_C0_P16(decoder, data + 1);
+			else
+				ccx_common_logging.debug_ftn(CCX_DMT_708, "[CEA-708] dtvcc_handle_C0: Not enough data for P16\n");
+		}
 		len = 3;
 	}
 	if (len == -1)
@@ -1633,6 +1654,9 @@ int dtvcc_handle_extended_char(dtvcc_service_decoder *decoder, unsigned char *da
 	ccx_common_logging.debug_ftn(CCX_DMT_708, "[CEA-708] In dtvcc_handle_extended_char, "
 						  "first data code: [%c], length: [%u]\n",
 				     data[0], data_length);
+	if (data_length < 1)
+		return 0;
+
 	unsigned char c = 0x20; // Default to space
 	unsigned char code = data[0];
 	if (/* data[i]>=0x00 && */ code <= 0x1F) // Comment to silence warning
@@ -1701,8 +1725,17 @@ void dtvcc_process_service_block(dtvcc_ctx *dtvcc,
 		}
 		else // Use extended set
 		{
-			used = dtvcc_handle_extended_char(decoder, data + i + 1, data_length - 1);
-			used++; // Since we had DTVCC_C0_EXT1
+			if (i + 1 >= data_length)
+			{
+				used = 1; // skip EXT1
+			}
+			else
+			{
+				used = dtvcc_handle_extended_char(decoder,
+								  data + i + 1,
+								  data_length - i - 1) +
+				       1;
+			}
 		}
 		i += used;
 	}
@@ -1754,6 +1787,12 @@ void dtvcc_process_current_packet(dtvcc_ctx *dtvcc, int len)
 
 		if (service_number == 7) // There is an extended header
 		{
+			if (pos + 1 >= dtvcc->current_packet + len)
+			{
+				ccx_common_logging.debug_ftn(CCX_DMT_708, "[CEA-708] dtvcc_process_current_packet: "
+									  "Truncated extended header, stopping.\n");
+				break;
+			}
 			pos++;
 			service_number = (pos[0] & 0x3F); // 6 more significant bits
 			// printf ("Extended header: Service number: [%d]\n",service_number);

src/lib_ccx/ccx_decoders_common.c

@@ -224,7 +224,12 @@ int do_cb(struct lib_cc_decode *ctx, unsigned char *cc_block, struct cc_subtitle
 void dinit_cc_decode(struct lib_cc_decode **ctx)
 {
 	struct lib_cc_decode *lctx = *ctx;
+#ifndef DISABLE_RUST
+	ccxr_dtvcc_free(lctx->dtvcc_rust);
+	lctx->dtvcc_rust = NULL;
+#else
 	dtvcc_free(&lctx->dtvcc);
+#endif
 	dinit_avc(&lctx->avc_ctx);
 	ccx_decoder_608_dinit_library(&lctx->context_cc608_field_1);
 	ccx_decoder_608_dinit_library(&lctx->context_cc608_field_2);
@@ -294,10 +299,16 @@ struct lib_cc_decode *init_cc_decode(struct ccx_decoders_common_settings_t *sett
 	ctx->no_rollup = setting->no_rollup;
 	ctx->noscte20 = setting->noscte20;
 
+#ifndef DISABLE_RUST
+	ctx->dtvcc_rust = ccxr_dtvcc_init(setting->settings_dtvcc);
+	ctx->dtvcc = NULL; // Not used when Rust is enabled
+#else
 	ctx->dtvcc = dtvcc_init(setting->settings_dtvcc);
 	if (!ctx->dtvcc)
 		fatal(EXIT_NOT_ENOUGH_MEMORY, "In init_cc_decode: Out of memory initializing dtvcc.");
 	ctx->dtvcc->is_active = setting->settings_dtvcc->enabled;
+	ctx->dtvcc_rust = NULL;
+#endif
 
 	if (setting->codec == CCX_CODEC_ATSC_CC)
 	{
@@ -477,6 +488,13 @@ void flush_cc_decode(struct lib_cc_decode *ctx, struct cc_subtitle *sub)
 			}
 		}
 	}
+#ifndef DISABLE_RUST
+	if (ccxr_dtvcc_is_active(ctx->dtvcc_rust))
+	{
+		ctx->current_field = 3;
+		ccxr_flush_active_decoders(ctx->dtvcc_rust);
+	}
+#else
 	if (ctx->dtvcc->is_active)
 	{
 		for (int i = 0; i < CCX_DTVCC_MAX_SERVICES; i++)
@@ -491,6 +509,7 @@ void flush_cc_decode(struct lib_cc_decode *ctx, struct cc_subtitle *sub)
 			}
 		}
 	}
+#endif
 }
 struct encoder_ctx *copy_encoder_context(struct encoder_ctx *ctx)
 {

src/lib_ccx/ccx_decoders_common.h

@@ -32,4 +32,10 @@ struct cc_subtitle *copy_subtitle(struct cc_subtitle *sub);
 void free_encoder_context(struct encoder_ctx *ctx);
 void free_decoder_context(struct lib_cc_decode *ctx);
 void free_subtitle(struct cc_subtitle *sub);
+
+#ifndef DISABLE_RUST
+// Rust FFI function to flush active CEA-708 service decoders
+extern void ccxr_flush_active_decoders(void *dtvcc_rust);
+#endif
+
 #endif

src/lib_ccx/ccx_decoders_isdb.c

@@ -724,16 +724,17 @@ static int parse_csi(ISDBSubContext *ctx, const uint8_t *buf, int len)
 	// Copy buf in arg
 	for (i = 0; *buf != 0x20; i++)
 	{
-		if (i >= (sizeof(arg)) + 1)
+		if (i >= sizeof(arg) - 1)
 		{
-			isdb_log("UnExpected CSI %d >= %d", sizeof(arg) + 1, i);
+			isdb_log("UnExpected CSI: too long");
 			break;
 		}
 		arg[i] = *buf;
 		buf++;
 	}
 	/* ignore terminating 0x20 character */
-	arg[i] = *buf++;
+	if (i < sizeof(arg))
+		arg[i] = *buf++;
 
 	switch (*buf)
 	{

src/lib_ccx/ccx_decoders_structs.h

@@ -208,6 +208,7 @@ struct lib_cc_decode
 	int false_pict_header;
 
 	dtvcc_ctx *dtvcc;
+	void *dtvcc_rust;  // Persistent Rust CEA-708 decoder context
 	int current_field;
 	// Analyse/use the picture information
 	int maxtref; // Use to remember the temporal reference number

src/lib_ccx/ccx_demuxer.c

@@ -285,6 +285,9 @@ static void ccx_demuxer_print_cfg(struct ccx_demuxer *ctx)
 		case CCX_SM_MXF:
 			mprint("MXF");
 			break;
+		case CCX_SM_SCC:
+			mprint("SCC");
+			break;
 #ifdef WTV_DEBUG
 		case CCX_SM_HEX_DUMP:
 			mprint("Hex");
@@ -348,7 +351,6 @@ struct ccx_demuxer *init_demuxer(void *parent, struct demuxer_cfg *cfg)
 		{
 			ctx->pinfo[i].got_important_streams_min_pts[j] = UINT64_MAX;
 		}
-		ctx->pinfo[i].initialized_ocr = 0;
 		ctx->pinfo[i].version = 0xFF; // Not real in a real stream since it's 5 bits. FF => Not initialized
 	}
 

src/lib_ccx/ccx_demuxer.h

@@ -35,7 +35,6 @@ struct program_info
 {
 	int pid;
 	int program_number;
-	int initialized_ocr; // Avoid initializing the OCR more than once
 	uint8_t analysed_PMT_once : 1;
 	uint8_t version;
 	uint8_t saved_section[1021];

src/lib_ccx/ccx_demuxer_mxf.c

@@ -75,12 +75,15 @@ enum MXFLocalTag
 void update_tid_lut(struct MXFContext *ctx, uint32_t track_id, uint8_t *track_number, struct ccx_rational edit_rate)
 {
 	int i;
+	debug("update_tid_lut: track_id=%u (0x%x), track_number=%02X%02X%02X%02X, cap_track_id=%u\n",
+	      track_id, track_id, track_number[0], track_number[1], track_number[2], track_number[3], ctx->cap_track_id);
 	// Update essence element key if we have track Id of caption
 	if (ctx->cap_track_id == track_id)
 	{
 		memcpy(ctx->cap_essence_key, mxf_essence_element_key, 12);
 		memcpy(ctx->cap_essence_key + 12, track_number, 4);
 		ctx->edit_rate = edit_rate;
+		debug("MXF: Found caption track, track_id=%u\n", track_id);
 	}
 
 	for (i = 0; i < ctx->nb_tracks; i++)
@@ -248,6 +251,7 @@ static int mxf_read_vanc_vbi_desc(struct ccx_demuxer *demux, uint64_t size)
 		{
 			case MXF_TAG_LTRACK_ID:
 				ctx->cap_track_id = buffered_get_be32(demux);
+				debug("MXF: VANC/VBI descriptor found, Linked Track ID = %u\n", ctx->cap_track_id);
 				update_cap_essence_key(ctx, ctx->cap_track_id);
 				break;
 			default:
@@ -304,6 +308,17 @@ static int mxf_read_cdp_data(struct ccx_demuxer *demux, int size, struct demuxer
 		log("Incomplete CDP packet\n");
 
 	ret = buffered_read(demux, data->buffer + data->len, cc_count * 3);
+	// Log first few bytes of cc_data for debugging
+	if (cc_count > 0)
+	{
+		unsigned char *cc_ptr = data->buffer + data->len;
+		debug("cc_data (first 6 triplets): ");
+		for (int j = 0; j < (cc_count < 6 ? cc_count : 6); j++)
+		{
+			debug("%02X%02X%02X ", cc_ptr[j * 3], cc_ptr[j * 3 + 1], cc_ptr[j * 3 + 2]);
+		}
+		debug("\n");
+	}
 	data->len += cc_count * 3;
 	demux->past += cc_count * 3;
 	len += ret;
@@ -361,7 +376,10 @@ static int mxf_read_vanc_data(struct ccx_demuxer *demux, uint64_t size, struct d
 	// uint8_t count; /* Currently unused */
 
 	if (size < 19)
+	{
+		debug("VANC data too small: %" PRIu64 " < 19\n", size);
 		goto error;
+	}
 
 	ret = buffered_read(demux, vanc_header, 16);
 
@@ -370,31 +388,39 @@ static int mxf_read_vanc_data(struct ccx_demuxer *demux, uint64_t size, struct d
 		return CCX_EOF;
 	len += ret;
 
+	debug("VANC header: num_packets=%d, line=0x%02x%02x, wrap_type=0x%02x, sample_config=0x%02x\n",
+	      vanc_header[1], vanc_header[2], vanc_header[3], vanc_header[4], vanc_header[5]);
+
 	for (int i = 0; i < vanc_header[1]; i++)
 	{
 		DID = buffered_get_byte(demux);
 		len++;
+		debug("VANC packet %d: DID=0x%02x\n", i, DID);
 		if (!(DID == 0x61 || DID == 0x80))
 		{
+			debug("DID 0x%02x not recognized as caption DID\n", DID);
 			goto error;
 		}
 
 		SDID = buffered_get_byte(demux);
 		len++;
+		debug("VANC packet %d: SDID=0x%02x\n", i, SDID);
 		if (SDID == 0x01)
 			debug("Caption Type 708\n");
 		else if (SDID == 0x02)
 			debug("Caption Type 608\n");
 
 		cdp_size = buffered_get_byte(demux);
+		debug("VANC packet %d: cdp_size=%d\n", i, cdp_size);
 		if (cdp_size + 19 > size)
 		{
-			debug("Incomplete cdp(%d) in anc data(%d)\n", cdp_size, size);
+			log("Incomplete cdp(%d) in anc data(%" PRIu64 ")\n", cdp_size, size);
 			goto error;
 		}
 		len++;
 
 		ret = mxf_read_cdp_data(demux, cdp_size, data);
+		debug("mxf_read_cdp_data returned %d, data->len=%d\n", ret, data->len);
 		len += ret;
 		// len += (3 + count + 4);
 	}
@@ -411,15 +437,33 @@ static int mxf_read_essence_element(struct ccx_demuxer *demux, uint64_t size, st
 	int ret;
 	struct MXFContext *ctx = demux->private_data;
 
+	debug("mxf_read_essence_element: ctx->type=%d (ANC=%d, VBI=%d), size=%" PRIu64 "\n",
+	      ctx->type, MXF_CT_ANC, MXF_CT_VBI, size);
+
 	if (ctx->type == MXF_CT_ANC)
 	{
 		data->bufferdatatype = CCX_RAW_TYPE;
 		ret = mxf_read_vanc_data(demux, size, data);
-		data->pts = ctx->cap_count;
+		debug("mxf_read_vanc_data returned %d, data->len=%d\n", ret, data->len);
+		// Calculate PTS in 90kHz units from frame count and edit rate
+		// edit_rate is frames per second (e.g., 25/1 for 25fps)
+		// PTS = frame_count * 90000 / fps = frame_count * 90000 * edit_rate.den / edit_rate.num
+		if (ctx->edit_rate.num > 0 && ctx->edit_rate.den > 0)
+		{
+			data->pts = (int64_t)ctx->cap_count * 90000 * ctx->edit_rate.den / ctx->edit_rate.num;
+		}
+		else
+		{
+			// Fallback to 25fps if edit_rate not set
+			data->pts = (int64_t)ctx->cap_count * 90000 / 25;
+		}
+		debug("Frame %d, PTS=%" PRId64 " (edit_rate=%d/%d)\n",
+		      ctx->cap_count, data->pts, ctx->edit_rate.num, ctx->edit_rate.den);
 		ctx->cap_count++;
 	}
 	else
 	{
+		debug("Skipping essence element (not ANC type)\n");
 		ret = buffered_skip(demux, size);
 		demux->past += ret;
 	}
@@ -514,6 +558,7 @@ static int read_packet(struct ccx_demuxer *demux, struct demuxer_data *data)
 	KLVPacket klv;
 	const MXFReadTableEntry *reader;
 	struct MXFContext *ctx = demux->private_data;
+	static int first_essence_logged = 0;
 	while ((ret = klv_read_packet(&klv, demux)) == 0)
 	{
 		debug("Key %02X%02X%02X%02X%02X%02X%02X%02X.%02X%02X%02X%02X%02X%02X%02X%02X size %" PRIu64 "\n",
@@ -523,8 +568,25 @@ static int read_packet(struct ccx_demuxer *demux, struct demuxer_data *data)
 		      klv.key[12], klv.key[13], klv.key[14], klv.key[15],
 		      klv.length);
 
+		// Check if this is an essence element key (first 12 bytes match)
+		if (IS_KLV_KEY(klv.key, mxf_essence_element_key) && !first_essence_logged)
+		{
+			debug("MXF: First essence element key: %02X%02X%02X%02X%02X%02X%02X%02X.%02X%02X%02X%02X%02X%02X%02X%02X\n",
+			      klv.key[0], klv.key[1], klv.key[2], klv.key[3],
+			      klv.key[4], klv.key[5], klv.key[6], klv.key[7],
+			      klv.key[8], klv.key[9], klv.key[10], klv.key[11],
+			      klv.key[12], klv.key[13], klv.key[14], klv.key[15]);
+			debug("MXF: cap_essence_key: %02X%02X%02X%02X%02X%02X%02X%02X.%02X%02X%02X%02X%02X%02X%02X%02X\n",
+			      ctx->cap_essence_key[0], ctx->cap_essence_key[1], ctx->cap_essence_key[2], ctx->cap_essence_key[3],
+			      ctx->cap_essence_key[4], ctx->cap_essence_key[5], ctx->cap_essence_key[6], ctx->cap_essence_key[7],
+			      ctx->cap_essence_key[8], ctx->cap_essence_key[9], ctx->cap_essence_key[10], ctx->cap_essence_key[11],
+			      ctx->cap_essence_key[12], ctx->cap_essence_key[13], ctx->cap_essence_key[14], ctx->cap_essence_key[15]);
+			first_essence_logged = 1;
+		}
+
 		if (IS_KLV_KEY(klv.key, ctx->cap_essence_key))
 		{
+			debug("MXF: Found ANC essence element, size=%" PRIu64 "\n", klv.length);
 			mxf_read_essence_element(demux, klv.length, data);
 			if (data->len > 0)
 				break;
@@ -566,8 +628,15 @@ int ccx_mxf_getmoredata(struct lib_ccx_ctx *ctx, struct demuxer_data **ppdata)
 		data->program_number = 1;
 		data->stream_pid = 1;
 		data->codec = CCX_CODEC_ATSC_CC;
-		data->tb.num = 1001;
-		data->tb.den = 30000;
+		// PTS is already calculated in 90kHz units by mxf_read_essence_element
+		data->tb.num = 1;
+		data->tb.den = 90000;
+
+		// Enable CEA-708 (DTVCC) decoder for MXF files with VANC captions
+		if (ctx->dec_global_setting && ctx->dec_global_setting->settings_dtvcc)
+		{
+			ctx->dec_global_setting->settings_dtvcc->enabled = 1;
+		}
 	}
 	else
 	{
@@ -576,6 +645,11 @@ int ccx_mxf_getmoredata(struct lib_ccx_ctx *ctx, struct demuxer_data **ppdata)
 
 	ret = read_packet(ctx->demux_ctx, data);
 
+	// Ensure timebase is 90kHz since PTS is calculated in 90kHz units
+	// CDP parsing may have set a frame-based timebase which would cause incorrect conversion
+	data->tb.num = 1;
+	data->tb.den = 90000;
+
 	return ret;
 }
 

src/lib_ccx/ccx_dtvcc.c

@@ -25,7 +25,7 @@ void dtvcc_process_data(struct dtvcc_ctx *dtvcc,
 			ccx_common_logging.debug_ftn(CCX_DMT_708, "[CEA-708] dtvcc_process_data: DTVCC Channel Packet Data\n");
 			if (cc_valid && dtvcc->is_current_packet_header_parsed)
 			{
-				if (dtvcc->current_packet_length > 253)
+				if (dtvcc->current_packet_length + 2 > CCX_DTVCC_MAX_PACKET_LENGTH)
 				{
 					ccx_common_logging.debug_ftn(CCX_DMT_708, "[CEA-708] dtvcc_process_data: "
 										  "Warning: Legal packet size exceeded (1), data not added.\n");
@@ -51,7 +51,7 @@ void dtvcc_process_data(struct dtvcc_ctx *dtvcc,
 			ccx_common_logging.debug_ftn(CCX_DMT_708, "[CEA-708] dtvcc_process_data: DTVCC Channel Packet Start\n");
 			if (cc_valid)
 			{
-				if (dtvcc->current_packet_length > CCX_DTVCC_MAX_PACKET_LENGTH - 1)
+				if (dtvcc->current_packet_length + 2 > CCX_DTVCC_MAX_PACKET_LENGTH)
 				{
 					ccx_common_logging.debug_ftn(CCX_DMT_708, "[CEA-708] dtvcc_process_data: "
 										  "Warning: Legal packet size exceeded (2), data not added.\n");

src/lib_ccx/ccx_dtvcc.h

@@ -10,4 +10,14 @@ void dtvcc_process_data(struct dtvcc_ctx *dtvcc,
 dtvcc_ctx *dtvcc_init(ccx_decoder_dtvcc_settings *opts);
 void dtvcc_free(dtvcc_ctx **);
 
+#ifndef DISABLE_RUST
+// Rust FFI functions for persistent CEA-708 decoder
+extern void *ccxr_dtvcc_init(struct ccx_decoder_dtvcc_settings *settings_dtvcc);
+extern void ccxr_dtvcc_free(void *dtvcc_rust);
+extern void ccxr_dtvcc_process_data(void *dtvcc_rust, const unsigned char cc_valid,
+	const unsigned char cc_type, const unsigned char data1, const unsigned char data2);
+extern int ccxr_dtvcc_is_active(void *dtvcc_rust);
+extern void ccxr_dtvcc_set_active(void *dtvcc_rust, int active);
+#endif
+
 #endif // CCEXTRACTOR_CCX_DTVCC_H

src/lib_ccx/ccx_encoders_common.c

@@ -775,6 +775,7 @@ struct encoder_ctx *init_encoder(struct encoder_cfg *opt)
 		return NULL;
 	}
 	ctx->in_fileformat = opt->in_format;
+	ctx->is_pal = (opt->in_format == 2);
 
 	/** used in case of SUB_EOD_MARKER */
 	ctx->prev_start = -1;
@@ -841,6 +842,9 @@ struct encoder_ctx *init_encoder(struct encoder_cfg *opt)
 	ctx->segment_last_key_frame = 0;
 	ctx->nospupngocr = opt->nospupngocr;
 	ctx->scc_framerate = opt->scc_framerate;
+	ctx->scc_accurate_timing = opt->scc_accurate_timing;
+	ctx->scc_last_transmission_end = 0;
+	ctx->scc_last_display_end = 0;
 
 	// Initialize teletext multi-page output arrays (issue #665)
 	ctx->tlt_out_count = 0;

src/lib_ccx/ccx_encoders_common.h

@@ -156,6 +156,11 @@ struct encoder_ctx
 	// SCC output framerate
 	int scc_framerate; // SCC output framerate: 0=29.97 (default), 1=24, 2=25, 3=30
 
+	// SCC accurate timing (issue #1120)
+	int scc_accurate_timing;	 // If 1, use bandwidth-aware timing for broadcast compliance
+	LLONG scc_last_transmission_end; // When last caption transmission ends (ms)
+	LLONG scc_last_display_end;	 // When last caption display ends (ms)
+
 	int new_sentence; // Capitalize next letter?
 
 	int program_number;
@@ -177,12 +182,12 @@ struct encoder_ctx
 
 	// OCR in SPUPNG
 	int nospupngocr;
+	int is_pal;
 
-	// Teletext multi-page output (issue #665)
 	struct ccx_s_write *tlt_out[MAX_TLT_PAGES_EXTRACT]; // Output files per teletext page
 	uint16_t tlt_out_pages[MAX_TLT_PAGES_EXTRACT];       // Page numbers for each output slot
 	unsigned int tlt_srt_counter[MAX_TLT_PAGES_EXTRACT]; // SRT counter per page
-	int tlt_out_count;                                    // Number of teletext output files
+	int tlt_out_count;				     // Number of teletext output files
 };
 
 #define INITIAL_ENC_BUFFER_CAPACITY 2048

src/lib_ccx/ccx_encoders_scc.c

@@ -10,6 +10,171 @@ unsigned char odd_parity(const unsigned char byte)
 	return byte | !(cc608_parity(byte) % 2) << 7;
 }
 
+/**
+ * SCC Accurate Timing Implementation (Issue #1120)
+ *
+ * EIA-608 bandwidth constraints:
+ * - 2 bytes per frame at 29.97 FPS (or configured frame rate)
+ * - Captions must be pre-loaded before display time
+ * - Each control code takes 2 bytes (sent twice for reliability = 4 bytes total)
+ * - Text characters take 1 byte each
+ */
+
+// Get frame rate value from scc_framerate setting
+// 0=29.97 (default), 1=24, 2=25, 3=30
+static float get_scc_fps_internal(int scc_framerate)
+{
+	switch (scc_framerate)
+	{
+		case 1:
+			return 24.0f;
+		case 2:
+			return 25.0f;
+		case 3:
+			return 30.0f;
+		default:
+			return 29.97f;
+	}
+}
+
+/**
+ * Calculate total bytes needed to transmit a caption
+ *
+ * Byte costs:
+ * - Control code (RCL, EOC, ENM, EDM): 2 bytes x 2 (sent twice) = 4 bytes
+ * - Preamble code: 2 bytes x 2 = 4 bytes
+ * - Tab offset: 2 bytes x 2 = 4 bytes
+ * - Mid-row code (color/style): 2 bytes x 2 = 4 bytes
+ * - Text character: 1 byte each
+ * - Padding: 1 byte if odd number of text bytes
+ */
+static unsigned int calculate_caption_bytes(const struct eia608_screen *data)
+{
+	unsigned int total_bytes = 0;
+
+	// RCL (Resume Caption Loading): 4 bytes
+	total_bytes += 4;
+
+	for (unsigned char row = 0; row < 15; ++row)
+	{
+		if (!data->row_used[row])
+			continue;
+
+		int first, last;
+		find_limit_characters(data->characters[row], &first, &last, CCX_DECODER_608_SCREEN_WIDTH);
+
+		if (first > last)
+			continue;
+
+		// Assume we need at least one preamble per row: 4 bytes
+		total_bytes += 4;
+
+		// Count characters on this row
+		unsigned int char_count = 0;
+		enum font_bits prev_font = FONT_REGULAR;
+		enum ccx_decoder_608_color_code prev_color = COL_WHITE;
+		int prev_col = -1;
+
+		for (int col = first; col <= last; ++col)
+		{
+			// Check if we need position codes
+			if (prev_col != col - 1 && prev_col != -1)
+			{
+				// Need preamble + possible tab offset: 4-8 bytes
+				total_bytes += 4;
+				if (col % 4 != 0)
+					total_bytes += 4; // Tab offset
+			}
+
+			// Check if we need mid-row style codes
+			if (data->fonts[row][col] != prev_font || data->colors[row][col] != prev_color)
+			{
+				total_bytes += 4; // Mid-row code
+				prev_font = data->fonts[row][col];
+				prev_color = data->colors[row][col];
+			}
+
+			// Text character
+			char_count++;
+			prev_col = col;
+		}
+
+		// Add text bytes (1 per character, rounded up to even)
+		total_bytes += char_count;
+		if (char_count % 2 == 1)
+			total_bytes++; // Padding
+	}
+
+	// EOC (End of Caption): 4 bytes
+	total_bytes += 4;
+
+	// ENM (Erase Non-displayed Memory): 4 bytes
+	total_bytes += 4;
+
+	return total_bytes;
+}
+
+/**
+ * Calculate the pre-roll start time for a caption
+ *
+ * @param display_time When the caption should appear on screen (ms)
+ * @param total_bytes Total bytes to transmit
+ * @param fps Frame rate
+ * @return Time to begin loading the caption (ms)
+ */
+static LLONG calculate_preroll_time(LLONG display_time, unsigned int total_bytes, float fps)
+{
+	// Calculate transmission time in milliseconds
+	// 2 bytes per frame, so frames_needed = (total_bytes + 1) / 2
+	float ms_per_frame = 1000.0f / fps;
+	unsigned int frames_needed = (total_bytes + 1) / 2;
+	LLONG transmission_time_ms = (LLONG)(frames_needed * ms_per_frame);
+
+	// Add 1 frame for EOC to be sent before display
+	LLONG one_frame_ms = (LLONG)ms_per_frame;
+
+	LLONG preroll_start = display_time - transmission_time_ms - one_frame_ms;
+
+	// Don't go negative
+	if (preroll_start < 0)
+		preroll_start = 0;
+
+	return preroll_start;
+}
+
+/**
+ * Check for collision with previous caption transmission and resolve it
+ *
+ * @param context Encoder context with timing state
+ * @param preroll_start Proposed pre-roll start time (will be modified if collision)
+ * @param display_time Caption display time (may be adjusted)
+ * @param fps Frame rate
+ * @return true if timing was adjusted due to collision
+ */
+static bool resolve_collision(struct encoder_ctx *context, LLONG *preroll_start,
+			      LLONG *display_time, float fps)
+{
+	// Check if our preroll would start before previous caption finishes transmitting
+	// This prevents bandwidth collision but allows visual overlap (like scc_tools)
+	// Visual overlap is fine - the EOC command swaps buffers atomically
+	if (context->scc_last_transmission_end > 0 &&
+	    *preroll_start < context->scc_last_transmission_end)
+	{
+		// Bandwidth collision detected - shift our caption forward
+		// Add 1 frame buffer to ensure no overlap
+		LLONG one_frame_ms = (LLONG)(1000.0f / fps);
+		LLONG new_preroll = context->scc_last_transmission_end + one_frame_ms;
+		LLONG shift = new_preroll - *preroll_start;
+
+		*preroll_start = new_preroll;
+		*display_time += shift;
+
+		return true;
+	}
+
+	return false;
+}
+
 struct control_code_info
 {
 	unsigned int byte1_odd;
@@ -689,8 +854,13 @@ void add_timestamp(const struct encoder_ctx *context, LLONG time, const bool dis
 
 	// SMPTE format - use configurable frame rate (issue #1191)
 	float fps = get_scc_fps(context->scc_framerate);
-	float frame = milli * fps / 1000;
-	fdprintf(context->out->fh, "%02u:%02u:%02u:%02.f\t", hour, minute, second, frame);
+	// Calculate frame number from milliseconds, ensuring it stays in valid range 0 to fps-1
+	// Use floor to avoid rounding up to fps (e.g., 29.97 -> 30 is invalid)
+	int max_frames = (int)fps;
+	int frame = (int)(milli * fps / 1000.0f);
+	if (frame >= max_frames)
+		frame = max_frames - 1; // Cap at max valid frame (e.g., 29 for 29.97fps)
+	fdprintf(context->out->fh, "%02u:%02u:%02u:%02d\t", hour, minute, second, frame);
 }
 
 void clear_screen(const struct encoder_ctx *context, LLONG end_time, const unsigned char channel, const bool disassemble)
@@ -710,8 +880,51 @@ int write_cc_buffer_as_scenarist(const struct eia608_screen *data, struct encode
 	unsigned char current_row = UINT8_MAX;
 	unsigned char current_column = UINT8_MAX;
 
-	// 1. Load the caption
-	add_timestamp(context, data->start_time, disassemble);
+	// Timing variables for accurate timing mode (issue #1120)
+	LLONG actual_start_time = data->start_time; // When caption should display
+	LLONG actual_end_time = data->end_time;	    // When caption should clear
+	LLONG preroll_start = data->start_time;	    // When to start loading (default: same as display)
+	float fps = get_scc_fps_internal(context->scc_framerate);
+	bool use_separate_display_time = false; // Whether to write EOC at separate timestamp
+
+	// If accurate timing is enabled, calculate pre-roll and handle collisions
+	if (context->scc_accurate_timing)
+	{
+		// Calculate total bytes needed for this caption
+		unsigned int total_bytes = calculate_caption_bytes(data);
+
+		// Calculate when we need to start loading
+		preroll_start = calculate_preroll_time(actual_start_time, total_bytes, fps);
+
+		// Check for collisions with previous caption and resolve
+		if (resolve_collision(context, &preroll_start, &actual_start_time, fps))
+		{
+			// Timing was adjusted due to collision
+			// Also adjust end time by the same amount
+			LLONG shift = actual_start_time - data->start_time;
+			actual_end_time = data->end_time + shift;
+		}
+
+		// Update timing state for next caption
+		float ms_per_frame = 1000.0f / fps;
+		unsigned int frames_needed = (total_bytes + 1) / 2;
+		LLONG transmission_time_ms = (LLONG)(frames_needed * ms_per_frame);
+		context->scc_last_transmission_end = preroll_start + transmission_time_ms;
+		context->scc_last_display_end = actual_end_time;
+
+		// Enable separate display timing (like scc_tools)
+		use_separate_display_time = true;
+
+		// 1. Load the caption at pre-roll time
+		add_timestamp(context, preroll_start, disassemble);
+	}
+	else
+	{
+		// Legacy mode: use original timing
+		// 1. Load the caption
+		add_timestamp(context, data->start_time, disassemble);
+	}
+
 	write_control_code(context->out->fh, data->channel, RCL, disassemble, &bytes_written);
 	for (uint8_t row = 0; row < 15; ++row)
 	{
@@ -794,12 +1007,26 @@ int write_cc_buffer_as_scenarist(const struct eia608_screen *data, struct encode
 		check_padding(context->out->fh, disassemble, &bytes_written);
 	}
 
-	// 2. Show the caption
+	// 2. Show the caption (EOC = End of Caption, makes it visible)
+	if (use_separate_display_time)
+	{
+		// For accurate timing: write display command at actual display time
+		// This matches scc_tools behavior where load and display are separate
+		add_timestamp(context, actual_start_time, disassemble);
+	}
 	write_control_code(context->out->fh, data->channel, EOC, disassemble, &bytes_written);
 	write_control_code(context->out->fh, data->channel, ENM, disassemble, &bytes_written);
 
-	// 3. Clear the caption
-	clear_screen(context, data->end_time, data->channel, disassemble);
+	// 3. Clear the caption at the end time
+	// In accurate timing mode, skip clear - the next caption's EOC will handle the transition
+	// This matches scc_tools behavior which doesn't write EDM between consecutive captions
+	if (!use_separate_display_time)
+	{
+		// Legacy mode: always write clear
+		clear_screen(context, actual_end_time, data->channel, disassemble);
+	}
+	// In accurate timing mode, scc_last_display_end is still tracked for reference
+	// but we don't write the clear command to avoid out-of-order timestamps
 
 	return 1;
 }

src/lib_ccx/ccx_encoders_spupng.c

@@ -251,6 +251,9 @@ void set_spupng_offset(void *ctx, int x, int y)
 	sp->xOffset = x;
 	sp->yOffset = y;
 }
+
+// Forward declaration for calculate_spupng_offsets
+static void calculate_spupng_offsets(struct spupng_t *sp, struct encoder_ctx *ctx);
 int save_spupng(const char *filename, uint8_t *bitmap, int w, int h,
 		png_color *palette, png_byte *alpha, int nb_color)
 {
@@ -384,7 +387,7 @@ int write_cc_bitmap_as_spupng(struct cc_subtitle *sub, struct encoder_ctx *conte
 	struct cc_bitmap *rect;
 	png_color *palette = NULL;
 	png_byte *alpha = NULL;
-	int wrote_opentag = 1;
+	int wrote_opentag = 0; // Track if we actually wrote the tag
 
 	x_pos = -1;
 	y_pos = -1;
@@ -395,13 +398,11 @@ int write_cc_bitmap_as_spupng(struct cc_subtitle *sub, struct encoder_ctx *conte
 		return 0;
 
 	inc_spupng_fileindex(sp);
-	write_sputag_open(sp, sub->start_time, sub->end_time - 1);
 
 	if (sub->nb_data == 0 && (sub->flags & SUB_EOD_MARKER))
 	{
 		context->prev_start = -1;
-		if (wrote_opentag)
-			write_sputag_close(sp);
+		// No subtitle data, skip writing
 		return 0;
 	}
 	rect = sub->data;
@@ -440,7 +441,13 @@ int write_cc_bitmap_as_spupng(struct cc_subtitle *sub, struct encoder_ctx *conte
 		}
 	}
 	filename = get_spupng_filename(sp);
-	set_spupng_offset(sp, x_pos, y_pos);
+
+	// Set image dimensions for offset calculation
+	sp->img_w = width;
+	sp->img_h = height;
+
+	// Calculate centered offsets based on screen size (PAL/NTSC)
+	calculate_spupng_offsets(sp, context);
 	if (sub->flags & SUB_EOD_MARKER)
 		context->prev_start = sub->start_time;
 	pbuf = (uint8_t *)malloc(width * height);
@@ -475,6 +482,15 @@ int write_cc_bitmap_as_spupng(struct cc_subtitle *sub, struct encoder_ctx *conte
 
 	/* TODO do rectangle wise, one color table should not be used for all rectangles */
 	mapclut_paletee(palette, alpha, (uint32_t *)rect[0].data1, rect[0].nb_colors);
+
+	// Save PNG file first
+	save_spupng(filename, pbuf, width, height, palette, alpha, rect[0].nb_colors);
+	freep(&pbuf);
+
+	// Write XML tag with calculated centered offsets
+	write_sputag_open(sp, sub->start_time, sub->end_time - 1);
+	wrote_opentag = 1; // Mark that we wrote the tag
+
 #ifdef ENABLE_OCR
 	if (!context->nospupngocr)
 	{
@@ -487,8 +503,6 @@ int write_cc_bitmap_as_spupng(struct cc_subtitle *sub, struct encoder_ctx *conte
 		}
 	}
 #endif
-	save_spupng(filename, pbuf, width, height, palette, alpha, rect[0].nb_colors);
-	freep(&pbuf);
 
 end:
 	if (wrote_opentag)
@@ -991,6 +1005,8 @@ int spupng_export_string2png(struct spupng_t *sp, char *str, FILE *output)
 	*/
 
 	// Save image
+	sp->img_w = canvas_width;
+	sp->img_h = canvas_height;
 	write_image(buffer, output, canvas_width, canvas_height);
 	free(tmp);
 	free(buffer);
@@ -1081,6 +1097,28 @@ int eia608_to_str(struct encoder_ctx *context, struct eia608_screen *data, char
 
 // string needs to be in UTF-8 encoding.
 // This function will take care of encoding.
+static void calculate_spupng_offsets(struct spupng_t *sp, struct encoder_ctx *ctx)
+{
+	int screen_w = 720;
+	int screen_h;
+
+	/* Teletext is always PAL */
+	if (ctx->in_fileformat == 2 || ctx->is_pal)
+	{
+		screen_h = 576;
+	}
+	else
+	{
+		screen_h = 480;
+	}
+
+	sp->xOffset = (screen_w - sp->img_w) / 2;
+	sp->yOffset = (screen_h - sp->img_h) / 2;
+
+	// SPU / DVD requires even yOffset (interlacing)
+	if (sp->yOffset & 1)
+		sp->yOffset++;
+}
 int spupng_write_string(struct spupng_t *sp, char *string, LLONG start_time, LLONG end_time,
 			struct encoder_ctx *context)
 {
@@ -1099,6 +1137,7 @@ int spupng_write_string(struct spupng_t *sp, char *string, LLONG start_time, LLO
 	}
 	// free(string_utf32);
 	fclose(sp->fppng);
+	calculate_spupng_offsets(sp, context);
 	write_sputag_open(sp, start_time, end_time);
 	write_spucomment(sp, string);
 	write_sputag_close(sp);

src/lib_ccx/ccx_encoders_structs.h

@@ -39,6 +39,8 @@ struct spupng_t
 	int fileIndex;
 	int xOffset;
 	int yOffset;
+	int img_w;
+	int img_h;
 };
 
 #endif

src/lib_ccx/dvb_subtitle_decoder.c

@@ -182,6 +182,7 @@ typedef struct DVBSubContext
 	LLONG time_out;
 #ifdef ENABLE_OCR
 	void *ocr_ctx;
+	int ocr_initialized; // Flag to track if OCR has been lazily initialized
 #endif
 	DVBSubRegion *region_list;
 	DVBSubCLUT *clut_list;
@@ -418,7 +419,7 @@ static void delete_regions(DVBSubContext *ctx)
  * @return DVB context kept as void* for abstraction
  *
  */
-void *dvbsub_init_decoder(struct dvb_config *cfg, int initialized_ocr)
+void *dvbsub_init_decoder(struct dvb_config *cfg)
 {
 	int i, r, g, b, a = 0;
 	DVBSubContext *ctx = (DVBSubContext *)malloc(sizeof(DVBSubContext));
@@ -442,8 +443,11 @@ void *dvbsub_init_decoder(struct dvb_config *cfg, int initialized_ocr)
 	}
 
 #ifdef ENABLE_OCR
-	if (!initialized_ocr)
-		ctx->ocr_ctx = init_ocr(ctx->lang_index);
+	// Lazy OCR initialization: don't init here, wait until a bitmap actually needs OCR
+	// This avoids ~10 second Tesseract startup overhead for files that have DVB streams
+	// but don't actually produce any bitmap subtitles (e.g., files with CEA-608 captions)
+	ctx->ocr_ctx = NULL;
+	ctx->ocr_initialized = 0;
 #endif
 	ctx->version = -1;
 
@@ -1702,7 +1706,13 @@ static int write_dvb_sub(struct lib_cc_decode *dec_ctx, struct cc_subtitle *sub)
 	// Perform OCR
 #ifdef ENABLE_OCR
 	char *ocr_str = NULL;
-	if (ctx->ocr_ctx)
+	// Lazy OCR initialization: only init when we actually have a bitmap to process
+	if (!ctx->ocr_initialized)
+	{
+		ctx->ocr_ctx = init_ocr(ctx->lang_index);
+		ctx->ocr_initialized = 1; // Mark as initialized even if init_ocr returns NULL
+	}
+	if (ctx->ocr_ctx && region)
 	{
 		int ret = ocr_rect(ctx->ocr_ctx, rect, &ocr_str, region->bgcolor, dec_ctx->ocr_quantmode);
 		if (ret >= 0)

src/lib_ccx/dvb_subtitle_decoder.h

@@ -42,7 +42,7 @@ extern "C"
 	 * @return DVB context kept as void* for abstraction
 	 *
 	 */
-	void *dvbsub_init_decoder(struct dvb_config *cfg, int initialized_ocr);
+	void *dvbsub_init_decoder(struct dvb_config *cfg);
 
 	int dvbsub_close_decoder(void **dvb_ctx);
 

src/lib_ccx/es_userdata.c

@@ -142,7 +142,7 @@ int user_data(struct encoder_ctx *enc_ctx, struct lib_cc_decode *dec_ctx, struct
 	{
 		if ((ud_header[1] & 0x7F) == 0x01)
 		{
-			unsigned char cc_data[3 * 31 + 1]; // Maximum cc_count is 31
+			unsigned char cc_data[3 * 32]; // Increased for safety margin, 31 is max count
 
 			dec_ctx->stat_scte20ccheaders++;
 			read_bytes(ustream, 2); // "03 01"
@@ -370,6 +370,7 @@ int user_data(struct encoder_ctx *enc_ctx, struct lib_cc_decode *dec_ctx, struct
 				dbg_print(CCX_DMT_PARSE, "%s", debug_608_to_ASC(dishdata, 0));
 				dbg_print(CCX_DMT_PARSE, "%s:\n", debug_608_to_ASC(dishdata + 3, 0));
 
+				dishdata[cc_count * 3] = 0xFF; // Ensure termination for store_hdcc
 				store_hdcc(enc_ctx, dec_ctx, dishdata, cc_count, dec_ctx->timing->current_tref, dec_ctx->timing->fts_now, sub);
 
 				// Ignore 4 (0x020A, followed by two unknown) bytes.
@@ -484,7 +485,10 @@ int user_data(struct encoder_ctx *enc_ctx, struct lib_cc_decode *dec_ctx, struct
 			mprint("MPEG:VBI: only support Luma line\n");
 
 		if (udatalen < 720)
-			mprint("MPEG:VBI: Minimum 720 bytes in luma line required\n");
+		{
+			mprint("MPEG:VBI: Minimum 720 bytes in luma line required, skipping truncated packet.\n");
+			return 1;
+		}
 
 		decode_vbi(dec_ctx, field, ustream->pos, 720, sub);
 		dbg_print(CCX_DMT_VERBOSE, "GXF (vbi line %d) user data:\n", line_nb);

src/lib_ccx/file_functions.c

@@ -66,6 +66,7 @@ void prepare_for_new_file(struct lib_ccx_ctx *ctx)
 {
 	// Init per file variables
 	ctx->last_reported_progress = -1;
+	ctx->min_global_timestamp_offset = -1; // -1 means not yet initialized
 	ctx->stat_numuserheaders = 0;
 	ctx->stat_dvdccheaders = 0;
 	ctx->stat_scte20ccheaders = 0;

src/lib_ccx/general_loop.c

@@ -18,6 +18,7 @@
 #include "ccx_gxf.h"
 #include "dvd_subtitle_decoder.h"
 #include "ccx_demuxer_mxf.h"
+#include "ccx_dtvcc.h"
 
 int end_of_file = 0; // End of file?
 
@@ -75,7 +76,7 @@ int ps_get_more_data(struct lib_ccx_ctx *ctx, struct demuxer_data **ppdata)
 				if (!ctx->demux_ctx->strangeheader)
 				{
 					mprint("\nNot a recognized header. Searching for next header.\n");
-					dump(CCX_DMT_GENERIC_NOTICES, nextheader, 6, 0, 0);
+					dump(CCX_DMT_PARSE, nextheader, 6, 0, 0);
 					// Only print the message once per loop / unrecognized header
 					ctx->demux_ctx->strangeheader = 1;
 				}
@@ -566,6 +567,104 @@ static size_t process_raw_for_mcc(struct encoder_ctx *enc_ctx, struct lib_cc_dec
 }
 
 // Raw file process
+// Parse raw CDP (Caption Distribution Packet) data
+// Returns number of bytes processed
+static size_t process_raw_cdp(struct encoder_ctx *enc_ctx, struct lib_cc_decode *dec_ctx,
+			      struct cc_subtitle *sub, unsigned char *buffer, size_t len)
+{
+	size_t pos = 0;
+	int cdp_count = 0;
+
+	while (pos + 10 < len) // Minimum CDP size
+	{
+		// Check for CDP identifier
+		if (buffer[pos] != 0x96 || buffer[pos + 1] != 0x69)
+		{
+			pos++;
+			continue;
+		}
+
+		unsigned char cdp_length = buffer[pos + 2];
+		if (pos + cdp_length > len)
+			break; // Incomplete CDP packet
+
+		unsigned char framerate_byte = buffer[pos + 3];
+		int framerate_code = framerate_byte >> 4;
+
+		// Skip to find cc_data section (0x72)
+		size_t cdp_pos = pos + 4; // After identifier, length, framerate
+		int cc_count = 0;
+		unsigned char *cc_data = NULL;
+
+		// Skip header sequence counter (2 bytes)
+		cdp_pos += 2;
+
+		// Look for cc_data section (0x72) within CDP
+		while (cdp_pos < pos + cdp_length - 4)
+		{
+			if (buffer[cdp_pos] == 0x72) // cc_data section
+			{
+				cc_count = buffer[cdp_pos + 1] & 0x1F;
+				cc_data = buffer + cdp_pos + 2;
+				break;
+			}
+			else if (buffer[cdp_pos] == 0x71) // time code section
+			{
+				cdp_pos += 5; // Skip time code section
+			}
+			else if (buffer[cdp_pos] == 0x73) // service info section
+			{
+				break; // Past cc_data
+			}
+			else if (buffer[cdp_pos] == 0x74) // footer
+			{
+				break;
+			}
+			else
+			{
+				cdp_pos++;
+			}
+		}
+
+		if (cc_count > 0 && cc_data != NULL)
+		{
+			// Calculate PTS based on CDP frame count and frame rate
+			static const int fps_table[] = {0, 24, 24, 25, 30, 30, 50, 60, 60};
+			int fps = (framerate_code < 9) ? fps_table[framerate_code] : 30;
+			LLONG pts = (LLONG)cdp_count * 90000 / fps;
+
+			// Set timing if not already set
+			if (dec_ctx->timing->pts_set == 0)
+			{
+				dec_ctx->timing->min_pts = pts;
+				dec_ctx->timing->pts_set = 2;
+				dec_ctx->timing->sync_pts = pts;
+			}
+			set_current_pts(dec_ctx->timing, pts);
+			set_fts(dec_ctx->timing);
+
+#ifndef DISABLE_RUST
+			// Enable DTVCC decoder for CEA-708 captions
+			if (dec_ctx->dtvcc_rust)
+			{
+				int is_active = ccxr_dtvcc_is_active(dec_ctx->dtvcc_rust);
+				if (!is_active)
+				{
+					ccxr_dtvcc_set_active(dec_ctx->dtvcc_rust, 1);
+				}
+			}
+#endif
+			// Process cc_data triplets through process_cc_data for 708 support
+			process_cc_data(enc_ctx, dec_ctx, cc_data, cc_count, sub);
+			cdp_count++;
+		}
+
+		pos += cdp_length;
+	}
+
+	return pos;
+}
+
 int raw_loop(struct lib_ccx_ctx *ctx)
 {
 	LLONG ret;
@@ -576,6 +675,7 @@ int raw_loop(struct lib_ccx_ctx *ctx)
 	int caps = 0;
 	int is_dvdraw = 0;     // Flag to track if this is DVD raw format
 	int is_scc = 0;	       // Flag to track if this is SCC format
+	int is_cdp = 0;	       // Flag to track if this is raw CDP format
 	int is_mcc_output = 0; // Flag for MCC output format
 
 	dec_ctx = update_decoder_list(ctx);
@@ -621,7 +721,15 @@ int raw_loop(struct lib_ccx_ctx *ctx)
 			mprint("Detected SCC (Scenarist Closed Caption) format\n");
 		}
 
-		if (is_mcc_output && !is_dvdraw && !is_scc)
+		// Check if this is raw CDP format (starts with 0x9669)
+		if (!is_cdp && !is_scc && !is_dvdraw && data->len >= 2 &&
+		    data->buffer[0] == 0x96 && data->buffer[1] == 0x69)
+		{
+			is_cdp = 1;
+			mprint("Detected raw CDP (Caption Distribution Packet) format\n");
+		}
+
+		if (is_mcc_output && !is_dvdraw && !is_scc && !is_cdp)
 		{
 			// For MCC output, encode raw data directly without decoding
 			// This preserves the original CEA-608 byte pairs in CDP format
@@ -639,6 +747,13 @@ int raw_loop(struct lib_ccx_ctx *ctx)
 			// Use Rust SCC implementation - handles timing internally via SMPTE timecodes
 			ret = ccxr_process_scc(dec_ctx, dec_sub, data->buffer, (unsigned int)data->len, ccx_options.scc_framerate);
 		}
+		else if (is_cdp)
+		{
+			// Process raw CDP packets (e.g., from SDI VANC capture)
+			ret = process_raw_cdp(enc_ctx, dec_ctx, dec_sub, data->buffer, data->len);
+			if (ret > 0)
+				caps = 1;
+		}
 		else
 		{
 			ret = process_raw(dec_ctx, dec_sub, data->buffer, data->len);
@@ -809,10 +924,6 @@ int process_data(struct encoder_ctx *enc_ctx, struct lib_cc_decode *dec_ctx, str
 			got = data_node->len;
 		}
 	}
-	else if (data_node->bufferdatatype == CCX_PRIVATE_MPEG2_CC)
-	{
-		got = data_node->len; // Do nothing. Still don't know how to process it
-	}
 	else if (data_node->bufferdatatype == CCX_RAW) // Raw two byte 608 data from DVR-MS/ASF
 	{
 		// The asf_get_more_data() loop sets current_pts when possible
@@ -865,7 +976,34 @@ int process_data(struct encoder_ctx *enc_ctx, struct lib_cc_decode *dec_ctx, str
 	}
 	else if (data_node->bufferdatatype == CCX_RAW_TYPE)
 	{
-		got = process_raw_with_field(dec_ctx, dec_sub, data_node->buffer, data_node->len);
+		// CCX_RAW_TYPE contains cc_data triplets (cc_type + 2 data bytes each)
+		// Used by MXF and GXF demuxers
+
+		// Initialize timing if not set (use caption PTS as reference)
+		if (dec_ctx->timing->pts_set == 0 && data_node->pts != CCX_NOPTS)
+		{
+			dec_ctx->timing->min_pts = data_node->pts;
+			dec_ctx->timing->pts_set = 2; // MinPtsSet
+			dec_ctx->timing->sync_pts = data_node->pts;
+			set_fts(dec_ctx->timing);
+		}
+
+#ifndef DISABLE_RUST
+		// Enable DTVCC decoder for CEA-708 captions from MXF/GXF
+		if (dec_ctx->dtvcc_rust)
+		{
+			int is_active = ccxr_dtvcc_is_active(dec_ctx->dtvcc_rust);
+			if (!is_active)
+			{
+				ccxr_dtvcc_set_active(dec_ctx->dtvcc_rust, 1);
+			}
+		}
+#endif
+
+		// Use process_cc_data to properly invoke DTVCC decoder for 708 captions
+		int cc_count = data_node->len / 3;
+		process_cc_data(enc_ctx, dec_ctx, data_node->buffer, cc_count, dec_sub);
+		got = data_node->len;
 	}
 	else if (data_node->bufferdatatype == CCX_ISDB_SUBTITLE)
 	{
@@ -1054,7 +1192,11 @@ int process_non_multiprogram_general_loop(struct lib_ccx_ctx *ctx,
 	cinfo = get_cinfo(ctx->demux_ctx, pid);
 	*enc_ctx = update_encoder_list_cinfo(ctx, cinfo);
 	*dec_ctx = update_decoder_list_cinfo(ctx, cinfo);
+#ifndef DISABLE_RUST
+	ccxr_dtvcc_set_encoder((*dec_ctx)->dtvcc_rust, *enc_ctx);
+#else
 	(*dec_ctx)->dtvcc->encoder = (void *)(*enc_ctx);
+#endif
 
 	if ((*dec_ctx)->timing->min_pts == 0x01FFFFFFFFLL) // if we didn't set the min_pts of the program
 	{
@@ -1278,7 +1420,11 @@ int general_loop(struct lib_ccx_ctx *ctx)
 
 				enc_ctx = update_encoder_list_cinfo(ctx, cinfo);
 				dec_ctx = update_decoder_list_cinfo(ctx, cinfo);
+#ifndef DISABLE_RUST
+				ccxr_dtvcc_set_encoder(dec_ctx->dtvcc_rust, enc_ctx);
+#else
 				dec_ctx->dtvcc->encoder = (void *)enc_ctx; // WARN: otherwise cea-708 will not work
+#endif
 
 				if (dec_ctx->timing->min_pts == 0x01FFFFFFFFLL) // if we didn't set the min_pts of the program
 				{
@@ -1362,7 +1508,24 @@ int general_loop(struct lib_ccx_ctx *ctx)
 		}
 		if (ctx->live_stream)
 		{
-			int cur_sec = (int)(get_fts(dec_ctx->timing, dec_ctx->current_field) / 1000);
+			LLONG t = get_fts(dec_ctx->timing, dec_ctx->current_field);
+			if (!t && ctx->demux_ctx->global_timestamp_inited)
+				t = ctx->demux_ctx->global_timestamp - ctx->demux_ctx->min_global_timestamp;
+			// Handle multi-program TS timing
+			if (ctx->demux_ctx->global_timestamp_inited)
+			{
+				LLONG offset = ctx->demux_ctx->global_timestamp - ctx->demux_ctx->min_global_timestamp;
+				if (ctx->min_global_timestamp_offset < 0 || offset < ctx->min_global_timestamp_offset)
+					ctx->min_global_timestamp_offset = offset;
+				// Only use timestamps from the program with the lowest base
+				if (offset - ctx->min_global_timestamp_offset < 60000)
+					t = offset - ctx->min_global_timestamp_offset;
+				else
+					t = ctx->min_global_timestamp_offset > 0 ? 0 : t;
+				if (t < 0)
+					t = 0;
+			}
+			int cur_sec = (int)(t / 1000);
 			int th = cur_sec / 10;
 			if (ctx->last_reported_progress != th)
 			{
@@ -1380,6 +1543,28 @@ int general_loop(struct lib_ccx_ctx *ctx)
 					LLONG t = get_fts(dec_ctx->timing, dec_ctx->current_field);
 					if (!t && ctx->demux_ctx->global_timestamp_inited)
 						t = ctx->demux_ctx->global_timestamp - ctx->demux_ctx->min_global_timestamp;
+					// For multi-program TS files, different programs can have different
+					// PCR bases (e.g., one at 25h, another at 23h). This causes the
+					// global_timestamp to jump between different bases, resulting in
+					// wildly different offset values. Track the minimum offset seen
+					// and only display times from the program with the lowest base.
+					if (ctx->demux_ctx->global_timestamp_inited)
+					{
+						LLONG offset = ctx->demux_ctx->global_timestamp - ctx->demux_ctx->min_global_timestamp;
+						// Track minimum offset (this is the PCR base of the program
+						// with the lowest timestamp, which represents true file time)
+						if (ctx->min_global_timestamp_offset < 0 || offset < ctx->min_global_timestamp_offset)
+							ctx->min_global_timestamp_offset = offset;
+						// Only use timestamps from the program with the lowest base.
+						// If current offset is significantly larger than minimum (by > 60s),
+						// it's from a program with a higher PCR base - use minimum instead.
+						if (offset - ctx->min_global_timestamp_offset < 60000)
+							t = offset - ctx->min_global_timestamp_offset;
+						else
+							t = ctx->min_global_timestamp_offset > 0 ? 0 : t; // fallback to minimum-based time
+						if (t < 0)
+							t = 0;
+					}
 					int cur_sec = (int)(t / 1000);
 					activity_progress(progress, cur_sec / 60, cur_sec % 60);
 					ctx->last_reported_progress = progress;
@@ -1488,7 +1673,11 @@ int rcwt_loop(struct lib_ccx_ctx *ctx)
 	}
 
 	dec_ctx = update_decoder_list(ctx);
+#ifndef DISABLE_RUST
+	ccxr_dtvcc_set_encoder(dec_ctx->dtvcc_rust, enc_ctx);
+#else
 	dec_ctx->dtvcc->encoder = (void *)enc_ctx; // WARN: otherwise cea-708 will not work
+#endif
 	if (parsebuf[6] == 0 && parsebuf[7] == 2)
 	{
 		dec_ctx->codec = CCX_CODEC_TELETEXT;

src/lib_ccx/lib_ccx.h

@@ -1,7 +1,7 @@
 #ifndef CCX_CCEXTRACTOR_H
 #define CCX_CCEXTRACTOR_H
 
-#define VERSION "0.96.3"
+#define VERSION "0.96.5"
 
 // Load common includes and constants for library usage
 #include "ccx_common_platform.h"
@@ -90,6 +90,7 @@ struct lib_ccx_ctx
 	LLONG total_past; // Only in binary concat mode
 
 	int last_reported_progress;
+	LLONG min_global_timestamp_offset; // Track minimum (global - min) for multi-program TS
 
 	/* Stats */
 	int stat_numuserheaders;
@@ -160,6 +161,7 @@ struct lib_ccx_ctx *init_libraries(struct ccx_s_options *opt);
 void dinit_libraries(struct lib_ccx_ctx **ctx);
 
 extern void ccxr_init_basic_logger();
+extern void ccxr_update_logger_target();
 
 // ccextractor.c
 void print_end_msg(void);
@@ -341,4 +343,9 @@ int process_non_multiprogram_general_loop(struct lib_ccx_ctx *ctx,
 void segment_output_file(struct lib_ccx_ctx *ctx, struct lib_cc_decode *dec_ctx);
 int decode_vbi(struct lib_cc_decode *dec_ctx, uint8_t field, unsigned char *buffer, size_t len, struct cc_subtitle *sub);
 
+#ifndef DISABLE_RUST
+// Rust FFI function to set encoder on persistent CEA-708 decoder
+void ccxr_dtvcc_set_encoder(void *dtvcc_rust, struct encoder_ctx *encoder);
+#endif
+
 #endif

src/lib_ccx/matroska.c

@@ -122,6 +122,8 @@ void parse_ebml(FILE *file)
 	{
 		code <<= 8;
 		code += mkv_read_byte(file);
+		if (feof(file))
+			break;
 		code_len++;
 
 		switch (code)
@@ -186,6 +188,8 @@ void parse_segment_info(FILE *file)
 	{
 		code <<= 8;
 		code += mkv_read_byte(file);
+		if (feof(file))
+			break;
 		code_len++;
 
 		switch (code)
@@ -484,6 +488,8 @@ void parse_segment_cluster_block_group(struct matroska_ctx *mkv_ctx, ULLONG clus
 	{
 		code <<= 8;
 		code += mkv_read_byte(file);
+		if (feof(file))
+			break;
 		code_len++;
 
 		switch (code)
@@ -612,6 +618,8 @@ void parse_segment_cluster(struct matroska_ctx *mkv_ctx)
 	{
 		code <<= 8;
 		code += mkv_read_byte(file);
+		if (feof(file))
+			break;
 		code_len++;
 
 		switch (code)
@@ -734,14 +742,24 @@ int process_avc_frame_mkv(struct matroska_ctx *mkv_ctx, struct matroska_avc_fram
 	{
 		uint32_t nal_length;
 
-		nal_length = bswap32(*(long *)&frame.data[i]);
+		if (i + nal_unit_size > frame.len)
+			break;
+
+		nal_length =
+		    ((uint32_t)frame.data[i] << 24) |
+		    ((uint32_t)frame.data[i + 1] << 16) |
+		    ((uint32_t)frame.data[i + 2] << 8) |
+		    (uint32_t)frame.data[i + 3];
+
 		i += nal_unit_size;
 
+		if (nal_length > frame.len - i)
+			break;
+
 		if (nal_length > 0)
-			do_NAL(enc_ctx, dec_ctx, (unsigned char *)&(frame.data[i]), nal_length, &mkv_ctx->dec_sub);
+			do_NAL(enc_ctx, dec_ctx, (unsigned char *)&frame.data[i], nal_length, &mkv_ctx->dec_sub);
 		i += nal_length;
 	} // outer for
-	assert(i == frame.len);
 
 	mkv_ctx->current_second = (int)(get_fts(dec_ctx->timing, dec_ctx->current_field) / 1000);
 
@@ -769,11 +787,22 @@ int process_hevc_frame_mkv(struct matroska_ctx *mkv_ctx, struct matroska_avc_fra
 	{
 		uint32_t nal_length;
 
-		nal_length = bswap32(*(long *)&frame.data[i]);
+		if (i + nal_unit_size > frame.len)
+			break;
+
+		nal_length =
+		    ((uint32_t)frame.data[i] << 24) |
+		    ((uint32_t)frame.data[i + 1] << 16) |
+		    ((uint32_t)frame.data[i + 2] << 8) |
+		    (uint32_t)frame.data[i + 3];
+
 		i += nal_unit_size;
 
+		if (nal_length > frame.len - i)
+			break;
+
 		if (nal_length > 0)
-			do_NAL(enc_ctx, dec_ctx, (unsigned char *)&(frame.data[i]), nal_length, &mkv_ctx->dec_sub);
+			do_NAL(enc_ctx, dec_ctx, (unsigned char *)&frame.data[i], nal_length, &mkv_ctx->dec_sub);
 		i += nal_length;
 	}
 
@@ -845,6 +874,8 @@ void parse_segment_track_entry(struct matroska_ctx *mkv_ctx)
 	{
 		code <<= 8;
 		code += mkv_read_byte(file);
+		if (feof(file))
+			break;
 		code_len++;
 
 		switch (code)
@@ -1173,7 +1204,7 @@ void parse_private_codec_data(struct matroska_ctx *mkv_ctx, char *codec_id_strin
 		memset((void *)&cnf, 0, sizeof(struct dvb_config));
 
 		parse_dvb_description(&cnf, codec_data, 8);
-		dec_ctx->private_data = dvbsub_init_decoder(&cnf, 0);
+		dec_ctx->private_data = dvbsub_init_decoder(&cnf);
 
 		free(codec_data);
 	}
@@ -1197,6 +1228,8 @@ void parse_segment_tracks(struct matroska_ctx *mkv_ctx)
 	{
 		code <<= 8;
 		code += mkv_read_byte(file);
+		if (feof(file))
+			break;
 		code_len++;
 
 		switch (code)
@@ -1241,6 +1274,8 @@ void parse_segment(struct matroska_ctx *mkv_ctx)
 	{
 		code <<= 8;
 		code += mkv_read_byte(file);
+		if (feof(file))
+			break;
 		code_len++;
 		switch (code)
 		{
@@ -1915,6 +1950,9 @@ void matroska_parse(struct matroska_ctx *mkv_ctx)
 	{
 		code <<= 8;
 		code += mkv_read_byte(file);
+		// Check for EOF after reading - feof() is only set after a failed read
+		if (feof(file))
+			break;
 		code_len++;
 
 		switch (code)

src/lib_ccx/mp4.c

@@ -749,7 +749,11 @@ static int process_clcp(struct lib_ccx_ctx *ctx, struct encoder_ctx *enc_ctx,
 					dbg_print(CCX_DMT_PARSE, "MP4-708: atom skipped (cc_type < 2)\n");
 					continue;
 				}
+#ifndef DISABLE_RUST
+				ccxr_dtvcc_process_data(dec_ctx->dtvcc_rust, cc_valid, cc_type, temp[2], temp[3]);
+#else
 				dtvcc_process_data(dec_ctx->dtvcc, (unsigned char *)temp);
+#endif
 				cb_708++;
 			}
 			if (ctx->write_format == CCX_OF_MCC)
@@ -887,10 +891,19 @@ int processmp4(struct lib_ccx_ctx *ctx, struct ccx_s_mp4Cfg *cfg, char *file)
 	if (enc_ctx)
 		enc_ctx->timing = dec_ctx->timing;
 
-	// WARN: otherwise cea-708 will not work
+		// WARN: otherwise cea-708 will not work
+#ifndef DISABLE_RUST
+	ccxr_dtvcc_set_encoder(dec_ctx->dtvcc_rust, enc_ctx);
+#else
 	dec_ctx->dtvcc->encoder = (void *)enc_ctx;
+#endif
 
 	memset(&dec_sub, 0, sizeof(dec_sub));
+	if (file == NULL)
+	{
+		mprint("Error: NULL file path provided to processmp4\n");
+		return -1;
+	}
 	mprint("Opening \'%s\': ", file);
 #ifdef MP4_DEBUG
 	gf_log_set_tool_level(GF_LOG_CONTAINER, GF_LOG_DEBUG);

src/lib_ccx/networking.c

@@ -103,7 +103,8 @@ int set_nonblocking(int fd);
 void connect_to_srv(const char *addr, const char *port, const char *cc_desc, const char *pwd)
 {
 #ifndef DISABLE_RUST
-	return ccxr_connect_to_srv(addr, port, cc_desc, pwd);
+	(void)ccxr_connect_to_srv(addr, port, cc_desc, pwd);
+	return;
 #endif
 	if (NULL == addr)
 	{
@@ -137,7 +138,8 @@ void connect_to_srv(const char *addr, const char *port, const char *cc_desc, con
 void net_send_header(const unsigned char *data, size_t len)
 {
 #ifndef DISABLE_RUST
-	return ccxr_net_send_header(data, len);
+	(void)ccxr_net_send_header(data, len);
+	return;
 #endif
 	assert(srv_sd > 0);
 
@@ -188,7 +190,8 @@ int net_send_cc(const unsigned char *data, int len, void *private_data, struct c
 void net_check_conn()
 {
 #ifndef DISABLE_RUST
-	return ccxr_net_check_conn();
+	ccxr_net_check_conn();
+	return;
 #endif
 	time_t now;
 	static time_t last_ping = 0;
@@ -252,7 +255,8 @@ void net_send_epg(
     const char *category)
 {
 #ifndef DISABLE_RUST
-	return ccxr_net_send_epg(start, stop, title, desc, lang, category);
+	(void)ccxr_net_send_epg(start, stop, title, desc, lang, category);
+	return;
 #endif
 	size_t st;
 	size_t sp;

src/lib_ccx/ocr.c

@@ -281,6 +281,13 @@ void *init_ocr(int lang_index)
 	// set PSM mode
 	TessBaseAPISetPageSegMode(ctx->api, ccx_options.psm);
 
+	// Set character blacklist to prevent common OCR errors (e.g. | vs I)
+	// These characters are rarely used in subtitles but often misrecognized
+	if (ccx_options.ocr_blacklist)
+	{
+		TessBaseAPISetVariable(ctx->api, "tessedit_char_blacklist", "|\\`_~");
+	}
+
 	free(pars_vec);
 	free(pars_values);
 
@@ -351,6 +358,176 @@ BOX *ignore_alpha_at_edge(png_byte *alpha, unsigned char *indata, int w, int h,
 	return cropWindow;
 }
 
+/**
+ * Structure to hold the vertical boundaries of a detected text line.
+ */
+struct line_bounds
+{
+	int start_y; // Top row of line (inclusive)
+	int end_y;   // Bottom row of line (inclusive)
+};
+
+/**
+ * Detects horizontal text line boundaries in a bitmap by finding rows of
+ * fully transparent pixels that separate lines of text.
+ *
+ * @param alpha     Palette alpha values (indexed by pixel value)
+ * @param indata    Bitmap pixel data (palette indices, w*h bytes)
+ * @param w         Image width
+ * @param h         Image height
+ * @param lines     Output: allocated array of line boundaries (caller must free)
+ * @param num_lines Output: number of lines found
+ * @param min_gap   Minimum consecutive transparent rows to count as line separator
+ * @return 0 on success, -1 on failure
+ */
+static int detect_text_lines(png_byte *alpha, unsigned char *indata,
+			     int w, int h,
+			     struct line_bounds **lines, int *num_lines,
+			     int min_gap)
+{
+	if (!alpha || !indata || !lines || !num_lines || w <= 0 || h <= 0)
+		return -1;
+
+	*lines = NULL;
+	*num_lines = 0;
+
+	// Allocate array to track which rows have visible content
+	int *row_has_content = (int *)malloc(h * sizeof(int));
+	if (!row_has_content)
+		return -1;
+
+	// Scan each row to determine if it has any visible (non-transparent) pixels
+	for (int i = 0; i < h; i++)
+	{
+		row_has_content[i] = 0;
+		for (int j = 0; j < w; j++)
+		{
+			int index = indata[i * w + j];
+			if (alpha[index] != 0)
+			{
+				row_has_content[i] = 1;
+				break; // Found visible pixel, no need to check rest of row
+			}
+		}
+	}
+
+	// Count lines by finding runs of content rows separated by gaps
+	int max_lines = (h / 2) + 1; // Conservative upper bound
+	struct line_bounds *temp_lines = (struct line_bounds *)malloc(max_lines * sizeof(struct line_bounds));
+	if (!temp_lines)
+	{
+		free(row_has_content);
+		return -1;
+	}
+
+	int line_count = 0;
+	int in_line = 0;
+	int line_start = 0;
+	int gap_count = 0;
+
+	for (int i = 0; i < h; i++)
+	{
+		if (row_has_content[i])
+		{
+			if (!in_line)
+			{
+				// Start of a new line
+				line_start = i;
+				in_line = 1;
+			}
+			gap_count = 0;
+		}
+		else
+		{
+			if (in_line)
+			{
+				gap_count++;
+				if (gap_count >= min_gap)
+				{
+					// End of line found (gap is large enough)
+					if (line_count < max_lines)
+					{
+						temp_lines[line_count].start_y = line_start;
+						temp_lines[line_count].end_y = i - gap_count;
+						line_count++;
+					}
+					in_line = 0;
+					gap_count = 0;
+				}
+			}
+		}
+	}
+
+	// Handle last line if we ended while still in a line
+	if (in_line && line_count < max_lines)
+	{
+		temp_lines[line_count].start_y = line_start;
+		// Find the last row with content
+		int last_content = h - 1;
+		while (last_content > line_start && !row_has_content[last_content])
+			last_content--;
+		temp_lines[line_count].end_y = last_content;
+		line_count++;
+	}
+
+	free(row_has_content);
+
+	if (line_count == 0)
+	{
+		free(temp_lines);
+		return -1;
+	}
+
+	// Shrink allocation to actual size
+	*lines = (struct line_bounds *)realloc(temp_lines, line_count * sizeof(struct line_bounds));
+	if (!*lines)
+	{
+		*lines = temp_lines; // Keep original if realloc fails
+	}
+	*num_lines = line_count;
+
+	return 0;
+}
+
+/**
+ * Performs OCR on a single text line image using PSM 7 (single line mode).
+ *
+ * @param ctx      OCR context (contains Tesseract API)
+ * @param line_pix Pre-processed PIX for single line (grayscale, inverted)
+ * @return Recognized text (caller must free with free()), or NULL on failure
+ */
+static char *ocr_single_line(struct ocrCtx *ctx, PIX *line_pix)
+{
+	if (!ctx || !ctx->api || !line_pix)
+		return NULL;
+
+	// Save current PSM
+	int saved_psm = TessBaseAPIGetPageSegMode(ctx->api);
+
+	// Set PSM 7 for single line recognition
+	TessBaseAPISetPageSegMode(ctx->api, 7); // PSM_SINGLE_LINE
+
+	// Perform OCR
+	TessBaseAPISetImage2(ctx->api, line_pix);
+	BOOL ret = TessBaseAPIRecognize(ctx->api, NULL);
+
+	char *text = NULL;
+	if (!ret)
+	{
+		char *tess_text = TessBaseAPIGetUTF8Text(ctx->api);
+		if (tess_text)
+		{
+			text = strdup(tess_text);
+			TessDeleteText(tess_text);
+		}
+	}
+
+	// Restore original PSM
+	TessBaseAPISetPageSegMode(ctx->api, saved_psm);
+
+	return text;
+}
+
 void debug_tesseract(struct ocrCtx *ctx, char *dump_path)
 {
 #ifdef OCR_DEBUG
@@ -397,6 +574,8 @@ char *ocr_bitmap(void *arg, png_color *palette, png_byte *alpha, unsigned char *
 	unsigned int *data, *ppixel;
 	BOOL tess_ret = FALSE;
 	struct ocrCtx *ctx = arg;
+	char *combined_text = NULL; // Used by line-split mode
+	size_t combined_len = 0;    // Used by line-split mode
 	pix = pixCreate(w, h, 32);
 	color_pix = pixCreate(w, h, 32);
 	if (pix == NULL || color_pix == NULL)
@@ -476,6 +655,98 @@ char *ocr_bitmap(void *arg, png_color *palette, png_byte *alpha, unsigned char *
 		return NULL;
 	}
 
+	// Line splitting mode: detect lines and OCR each separately with PSM 7
+	if (ccx_options.ocr_line_split && h > 30)
+	{
+		struct line_bounds *lines = NULL;
+		int num_lines = 0;
+
+		// Use min_gap of 3 rows to detect line boundaries
+		if (detect_text_lines(alpha, indata, w, h, &lines, &num_lines, 3) == 0 && num_lines > 1)
+		{
+			// Multiple lines detected - process each separately with PSM 7
+			// (combined_text and combined_len are declared at function scope)
+
+			for (int line_idx = 0; line_idx < num_lines; line_idx++)
+			{
+				int line_h = lines[line_idx].end_y - lines[line_idx].start_y + 1;
+				if (line_h <= 0)
+					continue;
+
+				// Extract line region from the grayscale image
+				BOX *line_box = boxCreate(0, lines[line_idx].start_y,
+							  pixGetWidth(cpix_gs), line_h);
+				PIX *line_pix_raw = pixClipRectangle(cpix_gs, line_box, NULL);
+				boxDestroy(&line_box);
+
+				if (line_pix_raw)
+				{
+					// Add white padding around the line (helps Tesseract with edge characters)
+					// The image is inverted (dark text on light bg), so add white (255) border
+					int padding = 10;
+					PIX *line_pix = pixAddBorderGeneral(line_pix_raw, padding, padding, padding, padding, 255);
+					pixDestroy(&line_pix_raw);
+					if (!line_pix)
+						continue;
+					char *line_text = ocr_single_line(ctx, line_pix);
+					pixDestroy(&line_pix);
+
+					if (line_text)
+					{
+						// Trim trailing whitespace from line
+						size_t line_len = strlen(line_text);
+						while (line_len > 0 && (line_text[line_len - 1] == '\n' ||
+									line_text[line_len - 1] == '\r' ||
+									line_text[line_len - 1] == ' '))
+						{
+							line_text[--line_len] = '\0';
+						}
+
+						if (line_len > 0)
+						{
+							// Append to combined result
+							size_t new_len = combined_len + line_len + 2; // +1 for newline, +1 for null
+							char *new_combined = (char *)realloc(combined_text, new_len);
+							if (new_combined)
+							{
+								combined_text = new_combined;
+								if (combined_len > 0)
+								{
+									combined_text[combined_len++] = '\n';
+								}
+								strcpy(combined_text + combined_len, line_text);
+								combined_len += line_len;
+							}
+						}
+						free(line_text);
+					}
+				}
+			}
+
+			free(lines);
+
+			if (combined_text && combined_len > 0)
+			{
+				// Successfully processed lines - skip whole-image OCR
+				// but continue to color detection below
+				goto line_split_color_detection;
+			}
+
+			// If we got here, line splitting didn't produce results
+			// Fall through to whole-image OCR
+			if (combined_text)
+				free(combined_text);
+			combined_text = NULL;
+		}
+		else
+		{
+			// Line detection failed or only 1 line - fall through to whole-image OCR
+			if (lines)
+				free(lines);
+		}
+	}
+
+	// Standard whole-image OCR path
 	TessBaseAPISetImage2(ctx->api, cpix_gs);
 	tess_ret = TessBaseAPIRecognize(ctx->api, NULL);
 	debug_tesseract(ctx, "./temp/");
@@ -518,6 +789,14 @@ char *ocr_bitmap(void *arg, png_color *palette, png_byte *alpha, unsigned char *
 		fatal(EXIT_NOT_ENOUGH_MEMORY, "In ocr_bitmap: Out of memory allocating text_out.");
 	}
 
+	// Jump target for line-split mode: use combined_text and continue with color detection
+	if (0)
+	{
+	line_split_color_detection:
+		text_out = combined_text;
+		combined_text = NULL; // Transfer ownership
+	}
+
 	// Begin color detection
 	// Using tlt_config.nofontcolor or ccx_options.nofontcolor (true when "--no-fontcolor" parameter used) to skip color detection if not required
 	// This is also skipped if --no-spupngocr is set since the OCR output won't be used anyway

src/lib_ccx/output.c

@@ -14,7 +14,19 @@ void dinit_write(struct ccx_s_write *wb)
 		return;
 	}
 	if (wb->fh > 0)
+	{
+		// Check if the file is empty before closing
+		off_t file_size = lseek(wb->fh, 0, SEEK_END);
 		close(wb->fh);
+
+		// Delete empty output files to avoid generating useless 0-byte files
+		// This commonly happens with -12 option when one field has no captions
+		if (file_size == 0 && wb->filename != NULL)
+		{
+			unlink(wb->filename);
+			mprint("Deleted empty output file: %s\n", wb->filename);
+		}
+	}
 	freep(&wb->filename);
 	freep(&wb->original_filename);
 	if (wb->with_semaphore && wb->semaphore_filename)

src/lib_ccx/params.c

@@ -401,6 +401,13 @@ void print_usage(void)
 	mprint("                       12    Sparse text with OSD.\n");
 	mprint("                       13    Raw line. Treat the image as a single text line,\n");
 	mprint("                       bypassing hacks that are Tesseract-specific.\n");
+	mprint("       --ocr-line-split: Split subtitle images into lines before OCR.\n");
+	mprint("                       Uses PSM 7 (single text line mode) for each line,\n");
+	mprint("                       which can improve accuracy for multi-line bitmap subtitles\n");
+	mprint("                       (VOBSUB, DVD, DVB).\n");
+	mprint("     --no-ocr-blacklist: Disable the OCR character blacklist. By default,\n");
+	mprint("                       CCExtractor blacklists characters like |, \\, `, _, ~\n");
+	mprint("                       that are commonly misrecognized (e.g. 'I' as '|').\n");
 	mprint("             --mkvlang: For MKV subtitles, select which language's caption\n");
 	mprint("                       stream will be processed. e.g. 'eng' for English.\n");
 	mprint("                       Language codes can be either the 3 letters bibliographic\n");

src/lib_ccx/telxcc.c

@@ -434,10 +434,21 @@ void remap_g0_charset(uint8_t c)
 {
 	if (c != primary_charset.current)
 	{
+		if (c >= 56)
+		{
+			fprintf(stderr, "- G0 Latin National Subset ID 0x%1x.%1x is out of bounds\n", (c >> 3), (c & 0x7));
+			return;
+		}
 		uint8_t m = G0_LATIN_NATIONAL_SUBSETS_MAP[c];
 		if (m == 0xff)
 		{
 			fprintf(stderr, "- G0 Latin National Subset ID 0x%1x.%1x is not implemented\n", (c >> 3), (c & 0x7));
+			return;
+		}
+		else if (m >= 14)
+		{
+			fprintf(stderr, "- G0 Latin National Subset index %d is out of bounds\n", m);
+			return;
 		}
 		else
 		{
@@ -1392,7 +1403,7 @@ int tlt_process_pes_packet(struct lib_cc_decode *dec_ctx, uint8_t *buffer, uint1
 	uint8_t pes_ext_flag;
 	// extension
 	uint32_t t = 0;
-	uint16_t i;
+	uint32_t i;
 	struct TeletextCtx *ctx = dec_ctx->private_data;
 	ctx->sentence_cap = sentence_cap;
 
@@ -1468,6 +1479,9 @@ int tlt_process_pes_packet(struct lib_cc_decode *dec_ctx, uint8_t *buffer, uint1
 	if (pes_packet_length > size)
 		pes_packet_length = size;
 
+	if (size < 9)
+		return CCX_OK;
+
 	// optional PES header marker bits (10.. ....)
 	if ((buffer[6] & 0xc0) == 0x80)
 	{
@@ -1480,8 +1494,16 @@ int tlt_process_pes_packet(struct lib_cc_decode *dec_ctx, uint8_t *buffer, uint1
 	{
 		if ((optional_pes_header_included == YES) && ((buffer[7] & 0x80) > 0))
 		{
-			ctx->using_pts = YES;
-			dbg_print(CCX_DMT_TELETEXT, "- PID 0xbd PTS available\n");
+			if (size < 14)
+			{
+				ctx->using_pts = NO;
+				dbg_print(CCX_DMT_TELETEXT, "- PID 0xbd PTS signaled but packet too short, using TS PCR\n");
+			}
+			else
+			{
+				ctx->using_pts = YES;
+				dbg_print(CCX_DMT_TELETEXT, "- PID 0xbd PTS available\n");
+			}
 		}
 		else
 		{
@@ -1554,11 +1576,17 @@ int tlt_process_pes_packet(struct lib_cc_decode *dec_ctx, uint8_t *buffer, uint1
 	if (optional_pes_header_included == YES)
 		i += 3 + optional_pes_header_length;
 
-	while (i <= pes_packet_length - 6)
+	while (i + 2 <= pes_packet_length)
 	{
 		uint8_t data_unit_id = buffer[i++];
 		uint8_t data_unit_len = buffer[i++];
 
+		if (i + data_unit_len > pes_packet_length)
+		{
+			dbg_print(CCX_DMT_TELETEXT, "- Teletext data unit length %u exceeds PES packet length, stopping.\n", data_unit_len);
+			break;
+		}
+
 		if ((data_unit_id == DATA_UNIT_EBU_TELETEXT_NONSUBTITLE) || (data_unit_id == DATA_UNIT_EBU_TELETEXT_SUBTITLE))
 		{
 			// teletext payload has always size 44 bytes

src/lib_ccx/ts_functions.c

@@ -6,6 +6,7 @@
 #include "dvb_subtitle_decoder.h"
 #include "ccx_decoders_isdb.h"
 #include "file_buffer.h"
+#include <inttypes.h>
 
 #ifdef DEBUG_SAVE_TS_PACKETS
 #include <sys/types.h>
@@ -153,12 +154,11 @@ enum ccx_bufferdata_type get_buffer_type(struct cap_info *cinfo)
 	{
 		return CCX_TELETEXT;
 	}
-	else if (cinfo->stream == CCX_STREAM_TYPE_PRIVATE_MPEG2 && cinfo->codec == CCX_CODEC_ATSC_CC)
-	{
-		return CCX_PRIVATE_MPEG2_CC;
-	}
-	else if (cinfo->stream == CCX_STREAM_TYPE_PRIVATE_USER_MPEG2 && cinfo->codec == CCX_CODEC_ATSC_CC)
+	else if ((cinfo->stream == CCX_STREAM_TYPE_PRIVATE_MPEG2 ||
+		  cinfo->stream == CCX_STREAM_TYPE_PRIVATE_USER_MPEG2) &&
+		 cinfo->codec == CCX_CODEC_ATSC_CC)
 	{
+		// ATSC CC can be in either private stream type - process both as PES
 		return CCX_PES;
 	}
 	else
@@ -567,17 +567,15 @@ int copy_capbuf_demux_data(struct ccx_demuxer *ctx, struct demuxer_data **data,
 	if (!cinfo->capbuf || !cinfo->capbuflen)
 		return -1;
 
-	if (ptr->bufferdatatype == CCX_PRIVATE_MPEG2_CC)
-	{
-		dump(CCX_DMT_GENERIC_NOTICES, cinfo->capbuf, cinfo->capbuflen, 0, 1);
-		// Bogus data, so we return something
-		ptr->buffer[ptr->len++] = 0xFA;
-		ptr->buffer[ptr->len++] = 0x80;
-		ptr->buffer[ptr->len++] = 0x80;
-		return CCX_OK;
-	}
 	if (cinfo->codec == CCX_CODEC_TELETEXT)
 	{
+		if (cinfo->capbuflen > BUFSIZE - ptr->len)
+		{
+			fatal(CCX_COMMON_EXIT_BUG_BUG,
+			      "Teletext packet (%" PRId64 ") larger than remaining buffer (%" PRId64 ").\n",
+			      cinfo->capbuflen, (int64_t)(BUFSIZE - ptr->len));
+		}
+
 		memcpy(ptr->buffer + ptr->len, cinfo->capbuf, cinfo->capbuflen);
 		ptr->len += cinfo->capbuflen;
 		return CCX_OK;
@@ -672,7 +670,6 @@ void cinfo_cremation(struct ccx_demuxer *ctx, struct demuxer_data **data)
 
 int copy_payload_to_capbuf(struct cap_info *cinfo, struct ts_payload *payload)
 {
-	int newcapbuflen;
 
 	if (cinfo->ignore == CCX_TRUE &&
 	    ((cinfo->stream != CCX_STREAM_TYPE_VIDEO_MPEG2 &&
@@ -698,17 +695,22 @@ int copy_payload_to_capbuf(struct cap_info *cinfo, struct ts_payload *payload)
 	}
 
 	// copy payload to capbuf
-	newcapbuflen = cinfo->capbuflen + payload->length;
-	if (newcapbuflen > cinfo->capbufsize)
+	if (payload->length > INT64_MAX - cinfo->capbuflen)
 	{
-		unsigned char *new_capbuf = (unsigned char *)realloc(cinfo->capbuf, newcapbuflen);
+		mprint("Error: capbuf size overflow\n");
+		return -1;
+	}
+	int64_t newcapbuflen = (int64_t)cinfo->capbuflen + payload->length;
+	if (newcapbuflen > (int64_t)cinfo->capbufsize)
+	{
+		unsigned char *new_capbuf = (unsigned char *)realloc(cinfo->capbuf, (size_t)newcapbuflen);
 		if (!new_capbuf)
 			return -1;
 		cinfo->capbuf = new_capbuf;
-		cinfo->capbufsize = newcapbuflen;
+		cinfo->capbufsize = newcapbuflen; // Note: capbufsize is int in struct cap_info
 	}
 	memcpy(cinfo->capbuf + cinfo->capbuflen, payload->start, payload->length);
-	cinfo->capbuflen = newcapbuflen;
+	cinfo->capbuflen = newcapbuflen; // Note: capbuflen is int in struct cap_info
 
 	return CCX_OK;
 }

src/lib_ccx/ts_functions.h

@@ -50,8 +50,8 @@ struct EPG_rating
 struct EPG_event
 {
 	uint32_t id;
-	char start_time_string[21]; //"YYYYMMDDHHMMSS +0000" = 20 chars
-	char end_time_string[21];
+	char start_time_string[74]; // "YYYYMMDDHHMMSS +0000" = 20 chars, 74 to silence compiler warning
+	char end_time_string[74];
 	uint8_t running_status;
 	uint8_t free_ca_mode;
 	char ISO_639_language_code[4];

src/lib_ccx/ts_info.c

@@ -173,7 +173,7 @@ static void *init_private_data(enum ccx_code_type codec)
 		case CCX_CODEC_TELETEXT:
 			return telxcc_init();
 		case CCX_CODEC_DVB:
-			return dvbsub_init_decoder(NULL, 0);
+			return dvbsub_init_decoder(NULL);
 		default:
 			return NULL;
 	}

src/lib_ccx/ts_tables.c

@@ -399,9 +399,7 @@ int parse_PMT(struct ccx_demuxer *ctx, unsigned char *buf, int len, struct progr
 					ret = parse_dvb_description(&cnf, es_info, desc_len);
 					if (ret < 0)
 						break;
-					ptr = dvbsub_init_decoder(&cnf, pinfo->initialized_ocr);
-					if (!pinfo->initialized_ocr)
-						pinfo->initialized_ocr = 1;
+					ptr = dvbsub_init_decoder(&cnf);
 					if (ptr == NULL)
 						break;
 					update_capinfo(ctx, elementary_PID, stream_type, CCX_CODEC_DVB, program_number, ptr);
@@ -413,9 +411,18 @@ int parse_PMT(struct ccx_demuxer *ctx, unsigned char *buf, int len, struct progr
 		{
 			// if this any generally used video stream tyoe get clashed with ATSC/SCTE standard
 			// then this code can go in some atsc flag
+			// Validate ES_info_length against buffer bounds to prevent heap overflow
+			if (i + 5 + ES_info_length > len)
+				break;
+
 			unsigned char *es_info = buf + i + 5;
-			for (desc_len = 0; (buf + i + 5 + ES_info_length) > es_info; es_info += desc_len)
+			unsigned char *es_info_end = buf + i + 5 + ES_info_length;
+			for (desc_len = 0; es_info_end > es_info; es_info += desc_len)
 			{
+				// Need at least 2 bytes for descriptor_tag and desc_len
+				if (es_info + 2 > es_info_end)
+					break;
+
 				enum ccx_mpeg_descriptor descriptor_tag = (enum ccx_mpeg_descriptor)(*es_info++);
 				int nb_service;
 				int is_608;
@@ -439,9 +446,18 @@ int parse_PMT(struct ccx_demuxer *ctx, unsigned char *buf, int len, struct progr
 
 		if (IS_FEASIBLE(ctx->codec, ctx->nocodec, CCX_CODEC_TELETEXT) && ES_info_length && stream_type == CCX_STREAM_TYPE_PRIVATE_MPEG2) // MPEG-2 Packetized Elementary Stream packets containing private data
 		{
+			// Validate ES_info_length against buffer bounds
+			if (i + 5 + ES_info_length > len)
+				continue;
+
 			unsigned char *es_info = buf + i + 5;
-			for (desc_len = 0; (buf + i + 5 + ES_info_length) - es_info; es_info += desc_len)
+			unsigned char *es_info_end = buf + i + 5 + ES_info_length;
+			for (desc_len = 0; es_info_end > es_info; es_info += desc_len)
 			{
+				// Need at least 2 bytes for descriptor_tag and desc_len
+				if (es_info + 2 > es_info_end)
+					break;
+
 				enum ccx_mpeg_descriptor descriptor_tag = (enum ccx_mpeg_descriptor)(*es_info++);
 				desc_len = (*es_info++);
 				if (!IS_VALID_TELETEXT_DESC(descriptor_tag))
@@ -576,6 +592,15 @@ void ts_buffer_psi_packet(struct ccx_demuxer *ctx)
 	else if (ccounter == ctx->PID_buffers[pid]->prev_ccounter + 1 || (ctx->PID_buffers[pid]->prev_ccounter == 0x0f && ccounter == 0))
 	{
 		ctx->PID_buffers[pid]->prev_ccounter = ccounter;
+		// Check for integer overflow and reasonable size limit (1MB)
+		if (ctx->PID_buffers[pid]->buffer_length > 1024 * 1024 ||
+		    payload_length > 1024 * 1024 ||
+		    ctx->PID_buffers[pid]->buffer_length + payload_length > 1024 * 1024)
+		{
+			dbg_print(CCX_DMT_GENERIC_NOTICES, "\rWarning: PSI buffer for PID %u exceeded reasonable limit (1MB), discarding.\n", pid);
+			return;
+		}
+
 		void *tmp = realloc(ctx->PID_buffers[pid]->buffer, ctx->PID_buffers[pid]->buffer_length + payload_length);
 		if (tmp == NULL)
 		{
@@ -614,6 +639,10 @@ int parse_PAT(struct ccx_demuxer *ctx)
 	payload_start = ctx->PID_buffers[0]->buffer + pointer_field + 1;
 	payload_length = ctx->PID_buffers[0]->buffer_length - (pointer_field + 1);
 
+	// Need at least 8 bytes to read header fields
+	if (payload_length < 8)
+		return 0;
+
 	section_number = payload_start[6];
 	last_section_number = payload_start[7];
 

src/lib_ccx/ts_tables_epg.c

@@ -125,7 +125,7 @@ void EPG_ATSC_calc_time(char *output, uint32_t time)
 	timeinfo.tm_hour = 0;
 	timeinfo.tm_isdst = -1;
 	mktime(&timeinfo);
-	snprintf(output, 21, "%02d%02d%02d%02d%02d%02d +0000", timeinfo.tm_year + 1900, timeinfo.tm_mon + 1, timeinfo.tm_mday, timeinfo.tm_hour, timeinfo.tm_min, timeinfo.tm_sec);
+	snprintf(output, 74, "%02d%02d%02d%02d%02d%02d +0000", timeinfo.tm_year + 1900, timeinfo.tm_mon + 1, timeinfo.tm_mday, timeinfo.tm_hour, timeinfo.tm_min, timeinfo.tm_sec);
 }
 
 // Fills event.start_time_string in XMLTV format with passed DVB time

src/lib_ccx/utility.c

@@ -179,16 +179,21 @@ void mprint(const char *fmt, ...)
 	if (!ccx_options.messages_target)
 		return;
 	va_start(args, fmt);
-	if (ccx_options.messages_target == CCX_MESSAGES_STDOUT)
+
+	FILE *target = (ccx_options.messages_target == CCX_MESSAGES_STDOUT) ? stdout : stderr;
+
+	if (fmt[0] == '\r')
 	{
-		vfprintf(stdout, fmt, args);
-		fflush(stdout);
-	}
-	else
-	{
-		vfprintf(stderr, fmt, args);
-		fflush(stderr);
+#ifndef _WIN32
+		fprintf(target, "\r\033[K"); // Clear the line first
+		fmt++;			     // Skip the '\r' so only the clean text gets printed next
+#endif
 	}
+	// Windows (legacy console) does not support ANSI sequences; fallback to standard \r; and vfprintf below handles it the old-fashioned way.
+
+	vfprintf(target, fmt, args);
+	fflush(target);
+
 	va_end(args);
 }
 

src/rust/Cargo.lock

@@ -161,6 +161,12 @@ version = "2.10.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "812e12b5285cc515a9c72a5c1d3b6d46a19dac5acfef5265968c166106e31dd3"
 
+[[package]]
+name = "by_address"
+version = "1.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "64fa3c856b712db6612c019f14756e64e4bcea13337a6b33b696333a9eaa2d06"
+
 [[package]]
 name = "camino"
 version = "1.2.1"
@@ -355,21 +361,18 @@ dependencies = [
  "windows-sys 0.61.2",
 ]
 
+[[package]]
+name = "fast-srgb8"
+version = "1.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dd2e7510819d6fbf51a5545c8f922716ecfb14df168a3242f7d33e0239efe6a1"
+
 [[package]]
 name = "fastrand"
 version = "2.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "37909eebbb50d72f9059c3b6d82c0463f2ff062c9e95845c43a6c9c0355411be"
 
-[[package]]
-name = "find-crate"
-version = "0.6.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "59a98bbaacea1c0eb6a0876280051b892eb73594fd90cf3b20e9c817029c57d2"
-dependencies = [
- "toml",
-]
-
 [[package]]
 name = "form_urlencoded"
 version = "1.2.2"
@@ -819,26 +822,26 @@ checksum = "384b8ab6d37215f3c5301a95a4accb5d64aa607f1fcb26a11b5303878451b4fe"
 
 [[package]]
 name = "palette"
-version = "0.6.1"
+version = "0.7.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8f9cd68f7112581033f157e56c77ac4a5538ec5836a2e39284e65bd7d7275e49"
+checksum = "4cbf71184cc5ecc2e4e1baccdb21026c20e5fc3dcf63028a086131b3ab00b6e6"
 dependencies = [
  "approx",
- "num-traits",
+ "fast-srgb8",
  "palette_derive",
  "phf",
 ]
 
 [[package]]
 name = "palette_derive"
-version = "0.6.1"
+version = "0.7.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "05eedf46a8e7c27f74af0c9cfcdb004ceca158cb1b918c6f68f8d7a549b3e427"
+checksum = "f5030daf005bface118c096f510ffb781fc28f9ab6a32ab224d8631be6851d30"
 dependencies = [
- "find-crate",
+ "by_address",
  "proc-macro2",
  "quote",
- "syn 1.0.109",
+ "syn 2.0.111",
 ]
 
 [[package]]
@@ -1436,15 +1439,6 @@ dependencies = [
  "zerovec",
 ]
 
-[[package]]
-name = "toml"
-version = "0.5.11"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f4f7f0dd8d50a853a531c426359045b1998f04219d88799810762cd4ad314234"
-dependencies = [
- "serde",
-]
-
 [[package]]
 name = "toml_datetime"
 version = "0.7.3"

src/rust/Cargo.toml

@@ -13,7 +13,7 @@ crate-type = ["staticlib"]
 [dependencies]
 log = "0.4.26"
 env_logger = "0.8.4"
-palette = "0.6.1"
+palette = "0.7"
 tesseract-sys = { version = "0.5.15", optional = true, default-features = false }
 leptonica-sys = { version = "= 0.4.6", optional = true, default-features = false }
 clap = { version = "4.5.31", features = ["derive"] }

src/rust/build.rs

@@ -84,7 +84,12 @@ fn main() {
     {
         builder = builder.clang_arg("-DENABLE_HARDSUBX");
 
-        // Add FFmpeg include paths for Mac
+        // Check FFMPEG_INCLUDE_DIR environment variable (works on all platforms)
+        if let Ok(ffmpeg_include) = env::var("FFMPEG_INCLUDE_DIR") {
+            builder = builder.clang_arg(format!("-I{}", ffmpeg_include));
+        }
+
+        // Add FFmpeg include paths for Mac (Homebrew)
         if cfg!(target_os = "macos") {
             // Try common Homebrew paths
             if std::path::Path::new("/opt/homebrew/include").exists() {
@@ -98,22 +103,23 @@ fn main() {
             if std::path::Path::new(cellar_ffmpeg).exists() {
                 // Find the FFmpeg version directory
                 if let Ok(entries) = std::fs::read_dir(cellar_ffmpeg) {
-                    for entry in entries {
-                        if let Ok(entry) = entry {
-                            let include_path = entry.path().join("include");
-                            if include_path.exists() {
-                                builder =
-                                    builder.clang_arg(format!("-I{}", include_path.display()));
-                                break;
-                            }
+                    for entry in entries.flatten() {
+                        let include_path = entry.path().join("include");
+                        if include_path.exists() {
+                            builder = builder.clang_arg(format!("-I{}", include_path.display()));
+                            break;
                         }
                     }
                 }
             }
+        }
 
-            // Also check environment variable
-            if let Ok(ffmpeg_include) = env::var("FFMPEG_INCLUDE_DIR") {
-                builder = builder.clang_arg(format!("-I{}", ffmpeg_include));
+        // On Linux, try pkg-config to find FFmpeg include paths
+        if cfg!(target_os = "linux") {
+            if let Ok(lib) = pkg_config::Config::new().probe("libavcodec") {
+                for path in lib.include_paths {
+                    builder = builder.clang_arg(format!("-I{}", path.display()));
+                }
             }
         }
     }

src/rust/lib_ccxr/src/common/constants.rs

@@ -147,7 +147,11 @@ pub const CCX_DECODER_608_SCREEN_WIDTH: usize = 32;
 pub const ONEPASS: usize = 120; // Bytes we can always look ahead without going out of limits
 pub const BUFSIZE: usize = 2048 * 1024 + ONEPASS; // 2 Mb plus the safety pass
 pub const MAX_CLOSED_CAPTION_DATA_PER_PICTURE: usize = 32;
-pub const EIA_708_BUFFER_LENGTH: usize = 2048; // TODO: Find out what the real limit is
+/// CEA-708 Service Input Buffer size.
+/// Specification minimum is 128 bytes per service, but we use 2048 bytes
+/// (16x the minimum) to provide a safety margin for buffer management.
+/// Reference: CEA-708-E Section 8.4.3 - Service Input Buffers
+pub const EIA_708_BUFFER_LENGTH: usize = 2048;
 pub const TS_PACKET_PAYLOAD_LENGTH: usize = 184; // From specs
 pub const SUBLINESIZE: usize = 2048; // Max. length of a .srt line - TODO: Get rid of this
 pub const STARTBYTESLENGTH: usize = 1024 * 1024;

src/rust/lib_ccxr/src/common/mkv_lang.rs

@@ -0,0 +1,385 @@
+//! MKV language filtering support.
+//!
+//! Matroska files support two language code formats:
+//! - ISO 639-2 (3-letter bibliographic codes): "eng", "fre", "chi"
+//! - BCP 47 / IETF language tags: "en-US", "fr-CA", "zh-Hans"
+//!
+//! This module provides [`MkvLangFilter`] for parsing and matching language codes.
+
+use std::fmt;
+use std::str::FromStr;
+
+/// A filter for matching MKV track languages.
+///
+/// Supports comma-separated lists of language codes in either:
+/// - ISO 639-2 format (3-letter codes like "eng", "fre")
+/// - BCP 47 format (tags like "en-US", "fr-CA", "zh-Hans")
+///
+/// # Examples
+///
+/// ```
+/// use lib_ccxr::common::MkvLangFilter;
+///
+/// // Single language
+/// let filter: MkvLangFilter = "eng".parse().unwrap();
+/// assert!(filter.matches("eng", None));
+///
+/// // Multiple languages
+/// let filter: MkvLangFilter = "eng,fre,chi".parse().unwrap();
+/// assert!(filter.matches("fre", None));
+///
+/// // BCP 47 matching
+/// let filter: MkvLangFilter = "en-US,fr-CA".parse().unwrap();
+/// assert!(filter.matches("eng", Some("en-US")));
+/// ```
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct MkvLangFilter {
+    /// The original input string (used for C FFI)
+    raw: String,
+    /// Parsed and validated language codes
+    codes: Vec<LanguageCode>,
+}
+
+/// A single language code, either ISO 639-2 or BCP 47.
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct LanguageCode {
+    /// The normalized (lowercase) code
+    code: String,
+}
+
+/// Error type for invalid language codes.
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct InvalidLanguageCode {
+    /// The invalid code
+    pub code: String,
+    /// Description of what's wrong
+    pub reason: &'static str,
+}
+
+impl fmt::Display for InvalidLanguageCode {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        write!(f, "invalid language code '{}': {}", self.code, self.reason)
+    }
+}
+
+impl std::error::Error for InvalidLanguageCode {}
+
+impl LanguageCode {
+    /// Validates and creates a new language code.
+    ///
+    /// Accepts:
+    /// - ISO 639-2 codes: 3 ASCII letters (e.g., "eng", "fre")
+    /// - BCP 47 tags: primary language with optional subtags separated by hyphens
+    ///   (e.g., "en-US", "fr-CA", "zh-Hans-CN")
+    ///
+    /// # BCP 47 Structure
+    /// - Primary language: 2-3 letters
+    /// - Script (optional): 4 letters (e.g., "Hans", "Latn")
+    /// - Region (optional): 2 letters or 3 digits (e.g., "US", "419")
+    /// - Variant (optional): 5-8 alphanumeric characters
+    pub fn new(code: &str) -> Result<Self, InvalidLanguageCode> {
+        let code = code.trim();
+
+        if code.is_empty() {
+            return Err(InvalidLanguageCode {
+                code: code.to_string(),
+                reason: "empty language code",
+            });
+        }
+
+        // Check for valid characters (alphanumeric and hyphens only)
+        if !code.chars().all(|c| c.is_ascii_alphanumeric() || c == '-') {
+            return Err(InvalidLanguageCode {
+                code: code.to_string(),
+                reason: "must contain only ASCII letters, digits, and hyphens",
+            });
+        }
+
+        // Cannot start or end with hyphen
+        if code.starts_with('-') || code.ends_with('-') {
+            return Err(InvalidLanguageCode {
+                code: code.to_string(),
+                reason: "cannot start or end with hyphen",
+            });
+        }
+
+        // Cannot have consecutive hyphens
+        if code.contains("--") {
+            return Err(InvalidLanguageCode {
+                code: code.to_string(),
+                reason: "cannot have consecutive hyphens",
+            });
+        }
+
+        // Validate subtag structure
+        let subtags: Vec<&str> = code.split('-').collect();
+
+        // First subtag must be the primary language (2-3 letters)
+        let primary = subtags[0];
+        if primary.len() < 2 || primary.len() > 3 {
+            return Err(InvalidLanguageCode {
+                code: code.to_string(),
+                reason: "primary language subtag must be 2-3 letters",
+            });
+        }
+        if !primary.chars().all(|c| c.is_ascii_alphabetic()) {
+            return Err(InvalidLanguageCode {
+                code: code.to_string(),
+                reason: "primary language subtag must contain only letters",
+            });
+        }
+
+        // Validate subsequent subtags
+        for subtag in subtags.iter().skip(1) {
+            if subtag.is_empty() {
+                return Err(InvalidLanguageCode {
+                    code: code.to_string(),
+                    reason: "empty subtag",
+                });
+            }
+
+            let len = subtag.len();
+            let all_alpha = subtag.chars().all(|c| c.is_ascii_alphabetic());
+            let all_digit = subtag.chars().all(|c| c.is_ascii_digit());
+            let all_alnum = subtag.chars().all(|c| c.is_ascii_alphanumeric());
+
+            // Valid subtag types:
+            // - Script: 4 letters (e.g., "Hans")
+            // - Region: 2 letters or 3 digits (e.g., "US", "419")
+            // - Variant: 5-8 alphanumeric, or 4 starting with digit
+            // - Extension: single letter followed by more subtags
+            // - Private use: 'x' followed by 1-8 char subtags
+            let valid = match len {
+                1 => subtag.chars().all(|c| c.is_ascii_alphanumeric()), // Extension singleton
+                2 => all_alpha,                                         // Region (2 letters)
+                3 => all_alpha || all_digit,                            // 3 letters or 3 digits
+                4 => all_alpha || (subtag.chars().next().unwrap().is_ascii_digit() && all_alnum), // Script or variant starting with digit
+                5..=8 => all_alnum, // Variant
+                _ => false,
+            };
+
+            if !valid {
+                return Err(InvalidLanguageCode {
+                    code: code.to_string(),
+                    reason: "invalid subtag format",
+                });
+            }
+        }
+
+        Ok(Self {
+            code: code.to_lowercase(),
+        })
+    }
+
+    /// Returns the normalized (lowercase) code.
+    pub fn as_str(&self) -> &str {
+        &self.code
+    }
+
+    /// Checks if this code matches a track's language.
+    ///
+    /// Matching rules:
+    /// 1. Exact match (case-insensitive)
+    /// 2. Prefix match for BCP 47 (e.g., "en" matches "en-US")
+    pub fn matches(&self, iso639: &str, bcp47: Option<&str>) -> bool {
+        let iso639_lower = iso639.to_lowercase();
+        let bcp47_lower = bcp47.map(|s| s.to_lowercase());
+
+        // Exact match on ISO 639-2
+        if self.code == iso639_lower {
+            return true;
+        }
+
+        // Exact match on BCP 47
+        if let Some(ref bcp) = bcp47_lower {
+            if self.code == *bcp {
+                return true;
+            }
+        }
+
+        // Prefix match: "en" matches "en-US", "eng" matches track with bcp47 "en-US"
+        // The filter code could be a prefix of the track's BCP 47 tag
+        if let Some(ref bcp) = bcp47_lower {
+            if bcp.starts_with(&self.code) && bcp[self.code.len()..].starts_with('-') {
+                return true;
+            }
+            // Or the track's BCP 47 could be a prefix of the filter
+            if self.code.starts_with(bcp.as_str()) && self.code[bcp.len()..].starts_with('-') {
+                return true;
+            }
+        }
+
+        false
+    }
+}
+
+impl FromStr for LanguageCode {
+    type Err = InvalidLanguageCode;
+
+    fn from_str(s: &str) -> Result<Self, Self::Err> {
+        Self::new(s)
+    }
+}
+
+impl fmt::Display for LanguageCode {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        write!(f, "{}", self.code)
+    }
+}
+
+impl MkvLangFilter {
+    /// Creates a new filter from a comma-separated list of language codes.
+    pub fn new(input: &str) -> Result<Self, InvalidLanguageCode> {
+        let input = input.trim();
+        if input.is_empty() {
+            return Err(InvalidLanguageCode {
+                code: String::new(),
+                reason: "empty language filter",
+            });
+        }
+
+        let codes: Result<Vec<LanguageCode>, _> = input.split(',').map(LanguageCode::new).collect();
+
+        Ok(Self {
+            raw: input.to_string(),
+            codes: codes?,
+        })
+    }
+
+    /// Returns the raw input string (for C FFI compatibility).
+    pub fn as_raw_str(&self) -> &str {
+        &self.raw
+    }
+
+    /// Returns the parsed language codes.
+    pub fn codes(&self) -> &[LanguageCode] {
+        &self.codes
+    }
+
+    /// Checks if any of the filter's codes match a track's language.
+    ///
+    /// # Arguments
+    /// - `iso639`: The track's ISO 639-2 language code (e.g., "eng")
+    /// - `bcp47`: The track's BCP 47 language tag, if available (e.g., "en-US")
+    pub fn matches(&self, iso639: &str, bcp47: Option<&str>) -> bool {
+        self.codes.iter().any(|code| code.matches(iso639, bcp47))
+    }
+}
+
+impl FromStr for MkvLangFilter {
+    type Err = InvalidLanguageCode;
+
+    fn from_str(s: &str) -> Result<Self, Self::Err> {
+        Self::new(s)
+    }
+}
+
+impl fmt::Display for MkvLangFilter {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        write!(f, "{}", self.raw)
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_iso639_codes() {
+        // Valid 3-letter codes
+        assert!(LanguageCode::new("eng").is_ok());
+        assert!(LanguageCode::new("fre").is_ok());
+        assert!(LanguageCode::new("chi").is_ok());
+        assert!(LanguageCode::new("ENG").is_ok()); // Case insensitive
+
+        // 2-letter codes (ISO 639-1 style, valid in BCP 47)
+        assert!(LanguageCode::new("en").is_ok());
+        assert!(LanguageCode::new("fr").is_ok());
+    }
+
+    #[test]
+    fn test_bcp47_codes() {
+        // Language + region
+        assert!(LanguageCode::new("en-US").is_ok());
+        assert!(LanguageCode::new("fr-CA").is_ok());
+        assert!(LanguageCode::new("pt-BR").is_ok());
+
+        // Language + script
+        assert!(LanguageCode::new("zh-Hans").is_ok());
+        assert!(LanguageCode::new("zh-Hant").is_ok());
+        assert!(LanguageCode::new("sr-Latn").is_ok());
+
+        // Language + script + region
+        assert!(LanguageCode::new("zh-Hans-CN").is_ok());
+        assert!(LanguageCode::new("zh-Hant-TW").is_ok());
+
+        // UN M.49 numeric region codes
+        assert!(LanguageCode::new("es-419").is_ok()); // Latin America
+    }
+
+    #[test]
+    fn test_invalid_codes() {
+        // Too short
+        assert!(LanguageCode::new("a").is_err());
+
+        // Invalid characters
+        assert!(LanguageCode::new("en_US").is_err()); // Underscore not allowed
+        assert!(LanguageCode::new("en US").is_err()); // Space not allowed
+        assert!(LanguageCode::new("ça").is_err()); // Non-ASCII
+
+        // Invalid structure
+        assert!(LanguageCode::new("-en").is_err()); // Leading hyphen
+        assert!(LanguageCode::new("en-").is_err()); // Trailing hyphen
+        assert!(LanguageCode::new("en--US").is_err()); // Double hyphen
+
+        // Empty
+        assert!(LanguageCode::new("").is_err());
+    }
+
+    #[test]
+    fn test_filter_multiple_codes() {
+        let filter = MkvLangFilter::new("eng,fre,chi").unwrap();
+        assert_eq!(filter.codes().len(), 3);
+        assert!(filter.matches("eng", None));
+        assert!(filter.matches("fre", None));
+        assert!(filter.matches("chi", None));
+        assert!(!filter.matches("spa", None));
+    }
+
+    #[test]
+    fn test_filter_bcp47_matching() {
+        let filter = MkvLangFilter::new("en-US,fr-CA").unwrap();
+
+        // Exact BCP 47 match
+        assert!(filter.matches("eng", Some("en-US")));
+        assert!(filter.matches("fre", Some("fr-CA")));
+
+        // No match
+        assert!(!filter.matches("eng", Some("en-GB")));
+        assert!(!filter.matches("eng", None));
+    }
+
+    #[test]
+    fn test_filter_mixed_formats() {
+        let filter = MkvLangFilter::new("eng,fr-CA,zh-Hans").unwrap();
+
+        assert!(filter.matches("eng", None));
+        assert!(filter.matches("fre", Some("fr-CA")));
+        assert!(filter.matches("chi", Some("zh-Hans")));
+    }
+
+    #[test]
+    fn test_case_insensitivity() {
+        let filter = MkvLangFilter::new("ENG,FR-CA").unwrap();
+        assert!(filter.matches("eng", None));
+        assert!(filter.matches("ENG", None));
+        assert!(filter.matches("fre", Some("fr-ca")));
+        assert!(filter.matches("FRE", Some("FR-CA")));
+    }
+
+    #[test]
+    fn test_raw_string_preserved() {
+        let filter = MkvLangFilter::new("eng,fre").unwrap();
+        assert_eq!(filter.as_raw_str(), "eng,fre");
+    }
+}

src/rust/lib_ccxr/src/common/mod.rs

@@ -18,8 +18,10 @@
 
 mod bitstream;
 mod constants;
+mod mkv_lang;
 mod options;
 
 pub use bitstream::*;
 pub use constants::*;
+pub use mkv_lang::*;
 pub use options::*;

src/rust/lib_ccxr/src/common/options.rs

@@ -462,8 +462,13 @@ pub struct Options {
     /// (0 = no quantization at all, 1 = CCExtractor's internal,
     ///  2 = reduce distinct color count in image for faster results.)
     pub ocr_quantmode: u8,
-    /// The name of the language stream for MKV
-    pub mkvlang: Option<Language>,
+    /// If true, split images into lines before OCR (uses PSM 7 for better accuracy)
+    pub ocr_line_split: bool,
+    /// If true, use character blacklist to prevent common OCR errors (e.g. | vs I)
+    pub ocr_blacklist: bool,
+    /// Language filter for MKV subtitle tracks.
+    /// Accepts comma-separated ISO 639-2 codes (e.g., "eng,fre") or BCP 47 tags (e.g., "en-US,fr-CA").
+    pub mkvlang: Option<super::MkvLangFilter>,
     /// If true, the video stream will be processed even if we're using a different one for subtitles.
     pub analyze_video_stream: bool,
 
@@ -519,6 +524,8 @@ pub struct Options {
     pub segment_on_key_frames_only: bool,
     /// SCC input framerate: 0=29.97 (default), 1=24, 2=25, 3=30
     pub scc_framerate: i32,
+    /// SCC accurate timing (issue #1120): if true, use bandwidth-aware timing for broadcast compliance
+    pub scc_accurate_timing: bool,
     pub debug_mask: DebugMessageMask,
 
     #[cfg(feature = "with_libcurl")]
@@ -584,6 +591,8 @@ impl Default for Options {
             ocr_oem: -1,
             psm: 3,
             ocr_quantmode: 0, // No quantization - better OCR accuracy for DVB subtitles
+            ocr_line_split: false, // Don't split images into lines by default
+            ocr_blacklist: true, // Use character blacklist by default to prevent | vs I errors
             mkvlang: Default::default(),
             analyze_video_stream: Default::default(),
             hardsubx_ocr_mode: Default::default(),
@@ -620,7 +629,8 @@ impl Default for Options {
             multiprogram: Default::default(),
             out_interval: -1,
             segment_on_key_frames_only: Default::default(),
-            scc_framerate: 0, // 0 = 29.97fps (default)
+            scc_framerate: 0,           // 0 = 29.97fps (default)
+            scc_accurate_timing: false, // Off by default for backwards compatibility (issue #1120)
             debug_mask: DebugMessageMask::new(
                 DebugMessageFlag::GENERIC_NOTICE,
                 DebugMessageFlag::VERBOSE,

src/rust/lib_ccxr/src/net/target.rs

@@ -82,7 +82,6 @@ impl<'a> SendTarget<'a> {
                 "Unable to connect, address passed is null\n"
             );
         }
-        info!("Target address: {}\n", config.target_addr); // TODO remove this
         info!("Target port: {}\n", config.port.unwrap_or(DEFAULT_TCP_PORT));
         let tcp_stream = TcpStream::connect((
             config.target_addr,

src/rust/lib_ccxr/src/teletext.rs

@@ -1154,10 +1154,9 @@ impl<'a> TeletextContext<'a> {
                     }
 
                     if v >= 0x20 {
-                        let u = char::from_u32(v as u32).unwrap();
+                        let u = char::from_u32(v as u32).unwrap_or(char::REPLACEMENT_CHARACTER);
                         self.page_buffer_cur.get_or_insert("".into()).push(u);
                         if logger().expect("could not access logger").is_gui_mode() {
-                            // For now we just handle the easy stuff
                             eprint!("{u}");
                         }
                     }
@@ -1225,13 +1224,15 @@ impl<'a> TeletextContext<'a> {
                 }
             }
             _ => {
-                ans = Some(Subtitle::new_text(
-                    self.page_buffer_cur.take().unwrap().into(),
-                    self.page_buffer.show_timestamp,
-                    self.page_buffer.hide_timestamp + Timestamp::from_millis(1),
-                    None,
-                    "TLT".into(),
-                ));
+                if let Some(cur) = self.page_buffer_cur.take() {
+                    ans = Some(Subtitle::new_text(
+                        cur.into(),
+                        self.page_buffer.show_timestamp,
+                        self.page_buffer.hide_timestamp + Timestamp::from_millis(1),
+                        None,
+                        "TLT".into(),
+                    ));
+                }
             }
         }
 
@@ -1251,34 +1252,43 @@ impl<'a> TeletextContext<'a> {
         capitalization_list: &[String],
     ) {
         // variable names conform to ETS 300 706, chapter 7.1.2
-        let address = (decode_hamming_8_4(packet.address[1]).unwrap() << 4)
-            | decode_hamming_8_4(packet.address[0]).unwrap();
+        let Some(addr1) = decode_hamming_8_4(packet.address[1]) else {
+            return;
+        };
+        let Some(addr0) = decode_hamming_8_4(packet.address[0]) else {
+            return;
+        };
+        let address = (addr1 << 4) | addr0;
         let mut m = address & 0x7;
         if m == 0 {
             m = 8;
         }
         let y = (address >> 3) & 0x1f;
         let designation_code = if y > 25 {
-            decode_hamming_8_4(packet.data[0]).unwrap()
+            decode_hamming_8_4(packet.data[0]).unwrap_or(0x00)
         } else {
             0x00
         };
 
         if y == 0 {
             // CC map
-            let i = (decode_hamming_8_4(packet.data[1]).unwrap() << 4)
-                | decode_hamming_8_4(packet.data[0]).unwrap();
-            let flag_subtitle = (decode_hamming_8_4(packet.data[5]).unwrap() & 0x08) >> 3;
+            let h1 = decode_hamming_8_4(packet.data[1]).unwrap_or(0);
+            let h0 = decode_hamming_8_4(packet.data[0]).unwrap_or(0);
+            let i = (h1 << 4) | h0;
+
+            let flag_subtitle = (decode_hamming_8_4(packet.data[5]).unwrap_or(0) & 0x08) >> 3;
             self.cc_map[i as usize] |= flag_subtitle << (m - 1);
 
             let flag_subtitle = flag_subtitle != 0;
 
             if flag_subtitle && (i < 0xff) {
-                let mut thisp = ((m as u32) << 8)
-                    | ((decode_hamming_8_4(packet.data[1]).unwrap() as u32) << 4)
-                    | (decode_hamming_8_4(packet.data[0]).unwrap() as u32);
-                let t1 = format!("{thisp:x}"); // Example: 1928 -> 788
-                thisp = t1.parse().unwrap();
+                let h1 = decode_hamming_8_4(packet.data[1]).unwrap_or(0) as u32;
+                let h0 = decode_hamming_8_4(packet.data[0]).unwrap_or(0) as u32;
+                let mut thisp = ((m as u32) << 8) | (h1 << 4) | h0;
+
+                let t1 = format!("{thisp:x}");
+                // Fallback to original value if parsing fails to avoid panics on malformed BCD
+                thisp = t1.parse().unwrap_or(thisp);
                 if !self.seen_sub_page[thisp as usize] {
                     self.seen_sub_page[thisp as usize] = true;
                     info!(
@@ -1288,36 +1298,28 @@ impl<'a> TeletextContext<'a> {
                 }
             }
             if (self.config.page.get() == 0.into()) && flag_subtitle && (i < 0xff) {
-                self.config.page.replace(
-                    (((m as u16) << 8)
-                        | ((decode_hamming_8_4(packet.data[1]).unwrap() as u16) << 4)
-                        | (decode_hamming_8_4(packet.data[0]).unwrap() as u16))
-                        .into(),
-                );
+                let h1 = decode_hamming_8_4(packet.data[1]).unwrap_or(0) as u16;
+                let h0 = decode_hamming_8_4(packet.data[0]).unwrap_or(0) as u16;
+
+                self.config
+                    .page
+                    .replace((((m as u16) << 8) | (h1 << 4) | h0).into());
                 info!("- No teletext page specified, first received suitable page is {}, not guaranteed\n", self.config.page.get());
             }
 
             // Page number and control bits
-            let page_number: TeletextPageNumber = (((m as u16) << 8)
-                | ((decode_hamming_8_4(packet.data[1]).unwrap() as u16) << 4)
-                | (decode_hamming_8_4(packet.data[0]).unwrap() as u16))
-                .into();
-            let charset = ((decode_hamming_8_4(packet.data[7]).unwrap() & 0x08)
-                | (decode_hamming_8_4(packet.data[7]).unwrap() & 0x04)
-                | (decode_hamming_8_4(packet.data[7]).unwrap() & 0x02))
-                >> 1;
-            // let flag_suppress_header = decode_hamming_8_4(packet.data[6]).unwrap() & 0x01;
-            // let flag_inhibit_display = (decode_hamming_8_4(packet.data[6]).unwrap() & 0x08) >> 3;
+            let h1 = decode_hamming_8_4(packet.data[1]).unwrap_or(0) as u16;
+            let h0 = decode_hamming_8_4(packet.data[0]).unwrap_or(0) as u16;
+            let page_number: TeletextPageNumber = (((m as u16) << 8) | (h1 << 4) | h0).into();
 
+            let c7 = decode_hamming_8_4(packet.data[7]).unwrap_or(0);
+            let charset = (c7 & 0x08 | c7 & 0x04 | c7 & 0x02) >> 1;
             // ETS 300 706, chapter 9.3.1.3:
             // When set to '1' the service is designated to be in Serial mode and the transmission of a page is terminated
             // by the next page header with a different page number.
             // When set to '0' the service is designated to be in Parallel mode and the transmission of a page is terminated
             // by the next page header with a different page number but the same magazine number.
-            // The same setting shall be used for all page headers in the service.
-            // ETS 300 706, chapter 7.2.1: Page is terminated by and excludes the next page header packet
-            // having the same magazine address in parallel transmission mode, or any magazine address in serial transmission mode.
-            self.transmission_mode = if decode_hamming_8_4(packet.data[7]).unwrap() & 0x01 == 0 {
+            self.transmission_mode = if c7 & 0x01 == 0 {
                 TransmissionMode::Parallel
             } else {
                 TransmissionMode::Serial
@@ -1353,19 +1355,17 @@ impl<'a> TeletextContext<'a> {
 
             // Now we have the begining of page transmission; if there is page_buffer pending, process it
             if self.page_buffer.tainted {
-                // Convert telx to UCS-2 before processing
                 for yt in 1..=23 {
                     for it in 0..40 {
                         if self.page_buffer.text[yt][it] != 0x00
                             && !self.page_buffer.g2_char_present[yt][it]
                         {
-                            self.page_buffer.text[yt][it] = self
-                                .g0_charset
-                                .ucs2_char(self.page_buffer.text[yt][it].try_into().unwrap());
+                            if let Ok(c) = self.page_buffer.text[yt][it].try_into() {
+                                self.page_buffer.text[yt][it] = self.g0_charset.ucs2_char(c);
+                            }
                         }
                     }
                 }
-                // it would be nice, if subtitle hides on previous video frame, so we contract 40 ms (1 frame @25 fps)
                 self.page_buffer.hide_timestamp = timestamp - Timestamp::from_millis(40);
                 if self.page_buffer.hide_timestamp > timestamp {
                     self.page_buffer.hide_timestamp = Timestamp::from_millis(0);
@@ -1544,12 +1544,14 @@ impl<'a> TeletextContext<'a> {
                     info!("- Programme Identification Data = ");
                     for i in 20..40 {
                         let c = self.g0_charset.ucs2_char(packet.data[i]);
-                        // strip any control codes from PID, eg. TVP station
                         if c < 0x20 {
                             continue;
                         }
 
-                        info!("{}", char::from_u32(c as u32).unwrap());
+                        info!(
+                            "{}",
+                            char::from_u32(c as u32).unwrap_or(char::REPLACEMENT_CHARACTER)
+                        );
                     }
                     info!("\n");
 
@@ -1580,7 +1582,7 @@ impl<'a> TeletextContext<'a> {
 
                     info!(
                         "- Universal Time Co-ordinated = {}\n",
-                        t0.to_ctime().unwrap()
+                        t0.to_ctime().as_deref().unwrap_or("unknown")
                     );
 
                     debug!(msg_type = DebugMessageFlag::TELETEXT; "- Transmission mode = {:?}\n", self.transmission_mode);
@@ -1589,8 +1591,13 @@ impl<'a> TeletextContext<'a> {
                         && matches!(self.config.date_format, TimestampFormat::Date { .. })
                         && !self.config.noautotimeref
                     {
-                        info!("- Broadcast Service Data Packet received, resetting UTC referential value to {}\n", t0.to_ctime().unwrap());
-                        *UTC_REFVALUE.write().unwrap() = t as u64;
+                        info!(
+                            "- Broadcast Service Data Packet received, resetting UTC referential value to {}\n",
+                            t0.to_ctime().as_deref().unwrap_or("unknown")
+                        );
+                        if let Ok(mut lock) = UTC_REFVALUE.write() {
+                            *lock = t as u64;
+                        }
                         self.states.pts_initialized = false;
                     }
 
@@ -1610,15 +1617,14 @@ impl<'a> TeletextContext<'a> {
             if let Some(subtitles) = subtitles {
                 // output any pending close caption
                 if self.page_buffer.tainted {
-                    // Convert telx to UCS-2 before processing
                     for yt in 1..=23 {
                         for it in 0..40 {
                             if self.page_buffer.text[yt][it] != 0x00
                                 && !self.page_buffer.g2_char_present[yt][it]
                             {
-                                self.page_buffer.text[yt][it] = self
-                                    .g0_charset
-                                    .ucs2_char(self.page_buffer.text[yt][it].try_into().unwrap());
+                                if let Ok(c) = self.page_buffer.text[yt][it].try_into() {
+                                    self.page_buffer.text[yt][it] = self.g0_charset.ucs2_char(c);
+                                }
                             }
                         }
                     }

src/rust/lib_ccxr/src/time/units.rs

@@ -225,9 +225,6 @@ impl Timestamp {
         let m = millis / 60000 - 60 * h;
         let s = millis / 1000 - 3600 * h - 60 * m;
         let u = millis - 3600000 * h - 60000 * m - 1000 * s;
-        if h > 24 {
-            println!("{h}")
-        }
         Ok((h.try_into()?, m as u8, s as u8, u as u16))
     }
 

src/rust/lib_ccxr/src/util/log.rs

@@ -269,6 +269,11 @@ impl<'a> CCExtractorLogger {
         self.target
     }
 
+    /// Sets the target for logging messages.
+    pub fn set_target(&mut self, target: OutputTarget) {
+        self.target = target;
+    }
+
     /// Check if the messages are intercepted by GUI.
     pub fn is_gui_mode(&self) -> bool {
         self.gui_mode
@@ -276,8 +281,16 @@ impl<'a> CCExtractorLogger {
 
     fn print(&self, args: &Arguments<'a>) {
         match &self.target {
-            OutputTarget::Stdout => print!("{args}"),
-            OutputTarget::Stderr => eprint!("{args}"),
+            OutputTarget::Stdout => {
+                print!("{args}");
+                // Flush stdout to ensure output appears immediately, especially when
+                // mixing with C code that also writes to stdout
+                let _ = std::io::Write::flush(&mut std::io::stdout());
+            }
+            OutputTarget::Stderr => {
+                eprint!("{args}");
+                let _ = std::io::Write::flush(&mut std::io::stderr());
+            }
             OutputTarget::Quiet => {}
         }
     }

src/rust/src/args.rs

@@ -28,7 +28,7 @@ const BURNEDIN_SUBTITLE_EXTRACTION: &str = "Burned-in subtitle extraction";
 #[derive(Debug, Parser)]
 #[command(name = "CCExtractor")]
 #[command(author = "Carlos Fernandez Sanz, Volker Quetschke.")]
-#[command(version = "1.0")]
+#[command(version = "0.96.5")]
 #[command(about = "Teletext portions taken from Petr Kutalek's telxcc
 --------------------------------------------------------------------------
 Originally based on McPoodle's tools. Check his page for lots of information
@@ -227,7 +227,7 @@ pub struct Args {
     /// "all[EUC-KR]") and it will encode specified charset to
     /// UTF-8 using iconv. See iconv documentation to check if
     /// required encoding/charset is supported.
-    #[arg(long="service", value_name="services", verbatim_doc_comment, help_heading=OPTION_AFFECT_PROCESSED)]
+    #[arg(long="service", alias="svc", value_name="services", verbatim_doc_comment, help_heading=OPTION_AFFECT_PROCESSED)]
     pub cea708services: Option<String>,
     /// With the exception of McPoodle's raw format, which is just the closed
     /// caption data with no other info, CCExtractor can usually detect the
@@ -295,6 +295,13 @@ pub struct Args {
     /// Example: --scc-framerate 25
     #[arg(long="scc-framerate", verbatim_doc_comment, value_name="fps", help_heading=OPTIONS_AFFECTING_INPUT_FILES)]
     pub scc_framerate: Option<String>,
+    /// Enable bandwidth-aware timing for SCC output (issue #1120).
+    /// When enabled, captions are pre-loaded ahead of their display time
+    /// based on the EIA-608 transmission bandwidth (2 bytes/frame).
+    /// This ensures YouTube and broadcast compliance by preventing
+    /// caption collisions. Use this for professional SCC output.
+    #[arg(long="scc-accurate-timing", verbatim_doc_comment, help_heading=OPTIONS_AFFECTING_INPUT_FILES)]
+    pub scc_accurate_timing: bool,
     /// By default, ccextractor will process input files in
     /// sequence as if they were all one large file (i.e.
     /// split by a generic, non video-aware tool. If you
@@ -395,10 +402,10 @@ pub struct Args {
     /// reference to the received data. Use this parameter if
     /// you prefer your own reference. Note: Current this only
     /// affects Teletext in timed transcript with --datets.
-    #[arg(long, verbatim_doc_comment, help_heading=OPTIONS_AFFECTING_INPUT_FILES)]
+    #[arg(long, alias="noautotimeref", verbatim_doc_comment, help_heading=OPTIONS_AFFECTING_INPUT_FILES)]
     pub no_autotimeref: bool,
     /// Ignore SCTE-20 data if present.
-    #[arg(long, verbatim_doc_comment, help_heading=OPTIONS_AFFECTING_INPUT_FILES)]
+    #[arg(long, alias="noscte20", verbatim_doc_comment, help_heading=OPTIONS_AFFECTING_INPUT_FILES)]
     pub no_scte20: bool,
     /// Create a separate file for CSS instead of inline.
     #[arg(long, verbatim_doc_comment, help_heading=OPTIONS_AFFECTING_INPUT_FILES)]
@@ -453,7 +460,7 @@ pub struct Args {
     /// Do not append a BOM (Byte Order Mark) to output
     /// files. Note that this may break files when using
     /// Windows. This is the default in non-Windows builds.
-    #[arg(long, verbatim_doc_comment, conflicts_with="bom", help_heading=OUTPUT_AFFECTING_OUTPUT_FILES)]
+    #[arg(long, alias="nobom", verbatim_doc_comment, conflicts_with="bom", help_heading=OUTPUT_AFFECTING_OUTPUT_FILES)]
     pub no_bom: bool,
     /// Encode subtitles in Unicode instead of Latin-1.
     #[arg(long, verbatim_doc_comment, help_heading=OUTPUT_AFFECTING_OUTPUT_FILES)]
@@ -486,7 +493,7 @@ pub struct Args {
     pub defaultcolor: Option<String>,
     /// Sentence capitalization. Use if you hate
     /// ALL CAPS in subtitles.
-    #[arg(long, verbatim_doc_comment, help_heading=OUTPUT_AFFECTING_OUTPUT_FILES)]
+    #[arg(long, alias="sc", verbatim_doc_comment, help_heading=OUTPUT_AFFECTING_OUTPUT_FILES)]
     pub sentencecap: bool,
     /// Add the contents of 'file' to the list of words
     /// that must be capitalized. For example, if file
@@ -630,6 +637,18 @@ pub struct Args {
     /// bypassing hacks that are Tesseract-specific.
     #[arg(long, verbatim_doc_comment, value_name="mode", help_heading=OUTPUT_AFFECTING_OUTPUT_FILES)]
     pub psm: Option<u8>,
+    /// Split subtitle images into lines before OCR.
+    /// Uses PSM 7 (single text line mode) for each line,
+    /// which can improve accuracy for multi-line bitmap subtitles
+    /// (VOBSUB, DVD, DVB).
+    #[arg(long, verbatim_doc_comment, help_heading=OUTPUT_AFFECTING_OUTPUT_FILES)]
+    pub ocr_line_split: bool,
+    /// Disable the OCR character blacklist.
+    /// By default, CCExtractor blacklists characters like |, \, `, _
+    /// that are commonly misrecognized (e.g. 'I' as '|').
+    /// Use this flag to disable the blacklist.
+    #[arg(long, verbatim_doc_comment, help_heading=OUTPUT_AFFECTING_OUTPUT_FILES)]
+    pub no_ocr_blacklist: bool,
     /// For MKV subtitles, select which language's caption
     /// stream will be processed. e.g. 'eng' for English.
     /// Language codes can be either the 3 letters bibliographic
@@ -682,7 +701,7 @@ pub struct Args {
     /// If you hate the repeated lines caused by the roll-up
     /// emulation, you can have ccextractor write only one
     /// line at a time, getting rid of these repeated lines.
-    #[arg(long, verbatim_doc_comment, help_heading=OUTPUT_AFFECTING_BUFFERING)]
+    #[arg(long, alias="noru", verbatim_doc_comment, help_heading=OUTPUT_AFFECTING_BUFFERING)]
     pub no_rollup: bool,
     /// roll-up captions can consist of 2, 3 or 4 visible
     /// lines at any time (the number of lines is part of
@@ -811,10 +830,10 @@ pub struct Args {
     #[arg(long, verbatim_doc_comment, help_heading=OUTPUT_AFFECTING_DEBUG_DATA)]
     pub parsedebug: bool,
     /// Print Program Association Table dump.
-    #[arg(long="parsePAT", verbatim_doc_comment, help_heading=OUTPUT_AFFECTING_DEBUG_DATA)]
+    #[arg(long="parsePAT", alias="pat", verbatim_doc_comment, help_heading=OUTPUT_AFFECTING_DEBUG_DATA)]
     pub parse_pat: bool,
     /// Print Program Map Table dump.
-    #[arg(long="parsePMT", verbatim_doc_comment, help_heading=OUTPUT_AFFECTING_DEBUG_DATA)]
+    #[arg(long="parsePMT", alias="pmt", verbatim_doc_comment, help_heading=OUTPUT_AFFECTING_DEBUG_DATA)]
     pub parse_pmt: bool,
     /// Hex-dump defective TS packets.
     #[arg(long, verbatim_doc_comment, help_heading=OUTPUT_AFFECTING_DEBUG_DATA)]
@@ -849,7 +868,7 @@ pub struct Args {
     /// for video streams that have both teletext packets
     /// and CEA-608/708 packets (if teletext is processed
     /// then CEA-608/708 processing is disabled).
-    #[arg(long, verbatim_doc_comment, conflicts_with="teletext", help_heading=TELETEXT_OPTIONS)]
+    #[arg(long, alias="noteletext", verbatim_doc_comment, conflicts_with="teletext", help_heading=TELETEXT_OPTIONS)]
     pub no_teletext: bool,
     /// Use the passed format to customize the (Timed) Transcript
     /// output. The format must be like this: 1100100 (7 digits).
@@ -978,6 +997,8 @@ pub enum InFormat {
     Mkv,
     /// Material Exchange Format (MXF).
     Mxf,
+    /// Scenarist Closed Caption (SCC).
+    Scc,
     #[cfg(feature = "wtv_debug")]
     // For WTV Debug mode only
     Hex,

src/rust/src/avc/core.rs

@@ -8,7 +8,9 @@ use crate::{anchor_hdcc, current_fps, process_hdcc, store_hdcc, MPEG_CLOCK_FREQ}
 use lib_ccxr::common::AvcNalType;
 use lib_ccxr::util::log::DebugMessageFlag;
 use lib_ccxr::{debug, info};
-#[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
+#[cfg(target_arch = "x86")]
+use std::arch::x86::*;
+#[cfg(target_arch = "x86_64")]
 use std::arch::x86_64::*;
 use std::os::raw::c_void;
 use std::slice;
@@ -453,7 +455,7 @@ pub fn hex_dump(data: &[u8]) {
 
         // Print hex bytes
         for byte in chunk {
-            print!("{:02X} ", byte);
+            print!("{byte:02X} ");
         }
 
         // Pad if less than 16 bytes

src/rust/src/avc/mod.rs

@@ -21,6 +21,19 @@ pub unsafe extern "C" fn ccxr_process_avc(
         return 0;
     }
 
+    // In report-only mode (-out=report), enc_ctx is NULL because no encoder is created.
+    // Skip AVC processing in this case since we can't output captions without an encoder.
+    // Return the full buffer length to indicate we've "consumed" the data.
+    if enc_ctx.is_null() {
+        return avcbuflen;
+    }
+
+    // dec_ctx and sub should never be NULL in normal operation, but check defensively
+    if dec_ctx.is_null() || sub.is_null() {
+        info!("Warning: dec_ctx or sub is NULL in ccxr_process_avc");
+        return avcbuflen;
+    }
+
     // Create a safe slice from the raw pointer
     let avc_slice = std::slice::from_raw_parts_mut(avcbuf, avcbuflen);
 

src/rust/src/avc/sei.rs

@@ -50,7 +50,7 @@ pub fn sei_message(ctx: &mut AvcContextRust, seibuf: &[u8]) -> usize {
         return 0;
     }
 
-    let mut payload_type = 0;
+    let mut payload_type: u32 = 0;
     while seibuf_idx < seibuf.len() && seibuf[seibuf_idx] == 0xff {
         payload_type += 255;
         seibuf_idx += 1;
@@ -60,10 +60,10 @@ pub fn sei_message(ctx: &mut AvcContextRust, seibuf: &[u8]) -> usize {
         return seibuf_idx;
     }
 
-    payload_type += seibuf[seibuf_idx] as i32;
+    payload_type += seibuf[seibuf_idx] as u32;
     seibuf_idx += 1;
 
-    let mut payload_size = 0;
+    let mut payload_size: u32 = 0;
     while seibuf_idx < seibuf.len() && seibuf[seibuf_idx] == 0xff {
         payload_size += 255;
         seibuf_idx += 1;
@@ -73,7 +73,7 @@ pub fn sei_message(ctx: &mut AvcContextRust, seibuf: &[u8]) -> usize {
         return seibuf_idx;
     }
 
-    payload_size += seibuf[seibuf_idx] as i32;
+    payload_size += seibuf[seibuf_idx] as u32;
     seibuf_idx += 1;
 
     let mut broken = false;
@@ -226,12 +226,10 @@ pub fn user_data_registered_itu_t_t35(ctx: &mut AvcContextRust, userbuf: &[u8])
                         }
 
                         // Save the data and process once we know the sequence number
-                        if ((ctx.cc_count as usize + local_cc_count) * 3) + 1 > ctx.cc_databufsize {
+                        let required_size = ((ctx.cc_count as usize + local_cc_count) * 3) + 1;
+                        if required_size > ctx.cc_data.len() {
                             let new_size = ((ctx.cc_count as usize + local_cc_count) * 6) + 1;
-                            unsafe {
-                                ctx.cc_data.set_len(new_size);
-                            }
-                            ctx.cc_data.reserve(new_size);
+                            ctx.cc_data.resize(new_size, 0);
                             ctx.cc_databufsize = new_size;
                         }
 

src/rust/src/common.rs

@@ -18,6 +18,7 @@ use lib_ccxr::common::DtvccServiceCharset;
 use lib_ccxr::common::EncoderConfig;
 use lib_ccxr::common::EncodersTranscriptFormat;
 use lib_ccxr::common::Language;
+use lib_ccxr::common::MkvLangFilter;
 use lib_ccxr::common::Options;
 use lib_ccxr::common::OutputFormat;
 use lib_ccxr::common::SelectCodec;
@@ -181,9 +182,11 @@ pub unsafe fn copy_from_rust(ccx_s_options: *mut ccx_s_options, options: Options
     (*ccx_s_options).ocr_oem = options.ocr_oem as _;
     (*ccx_s_options).psm = options.psm as _;
     (*ccx_s_options).ocr_quantmode = options.ocr_quantmode as _;
-    if let Some(mkvlang) = options.mkvlang {
+    (*ccx_s_options).ocr_line_split = options.ocr_line_split as _;
+    (*ccx_s_options).ocr_blacklist = options.ocr_blacklist as _;
+    if let Some(ref mkvlang) = options.mkvlang {
         (*ccx_s_options).mkvlang =
-            replace_rust_c_string((*ccx_s_options).mkvlang, mkvlang.to_ctype().as_str());
+            replace_rust_c_string((*ccx_s_options).mkvlang, mkvlang.as_raw_str());
     }
     (*ccx_s_options).analyze_video_stream = options.analyze_video_stream as _;
     (*ccx_s_options).hardsubx_ocr_mode = options.hardsubx_ocr_mode.to_ctype();
@@ -209,11 +212,9 @@ pub unsafe fn copy_from_rust(ccx_s_options: *mut ccx_s_options, options: Options
             replace_rust_c_string((*ccx_s_options).udpaddr, &options.udpaddr.clone().unwrap());
     }
     (*ccx_s_options).udpport = options.udpport as _;
-    if options.tcpport.is_some() {
-        (*ccx_s_options).tcpport = replace_rust_c_string(
-            (*ccx_s_options).tcpport,
-            &options.tcpport.unwrap().to_string(),
-        );
+    if let Some(tcpport) = options.tcpport {
+        (*ccx_s_options).tcpport =
+            replace_rust_c_string((*ccx_s_options).tcpport, &tcpport.to_string());
     }
     if options.tcp_password.is_some() {
         (*ccx_s_options).tcp_password = replace_rust_c_string(
@@ -233,11 +234,9 @@ pub unsafe fn copy_from_rust(ccx_s_options: *mut ccx_s_options, options: Options
             &options.srv_addr.clone().unwrap(),
         );
     }
-    if options.srv_port.is_some() {
-        (*ccx_s_options).srv_port = replace_rust_c_string(
-            (*ccx_s_options).srv_port,
-            &options.srv_port.unwrap().to_string(),
-        );
+    if let Some(srv_port) = options.srv_port {
+        (*ccx_s_options).srv_port =
+            replace_rust_c_string((*ccx_s_options).srv_port, &srv_port.to_string());
     }
     (*ccx_s_options).noautotimeref = options.noautotimeref as _;
     (*ccx_s_options).input_source = options.input_source as _;
@@ -251,15 +250,12 @@ pub unsafe fn copy_from_rust(ccx_s_options: *mut ccx_s_options, options: Options
     // Subsequent calls from ccxr_demuxer_open/close should NOT modify inputfile because
     // C code holds references to those strings throughout processing.
     // Freeing them would cause use-after-free and double-free errors.
-    if options.inputfile.is_some() && (*ccx_s_options).inputfile.is_null() {
-        (*ccx_s_options).inputfile = string_to_c_chars(options.inputfile.clone().unwrap());
-        (*ccx_s_options).num_input_files = options
-            .inputfile
-            .as_ref()
-            .unwrap()
-            .iter()
-            .filter(|s| !s.is_empty())
-            .count() as _;
+    if let Some(ref inputfile) = options.inputfile {
+        if (*ccx_s_options).inputfile.is_null() {
+            (*ccx_s_options).inputfile = string_to_c_chars(inputfile.clone());
+            (*ccx_s_options).num_input_files =
+                inputfile.iter().filter(|s| !s.is_empty()).count() as _;
+        }
     }
     (*ccx_s_options).demux_cfg = options.demux_cfg.to_ctype();
     // Only set enc_cfg on the first call (when output_filename is null).
@@ -278,6 +274,7 @@ pub unsafe fn copy_from_rust(ccx_s_options: *mut ccx_s_options, options: Options
     (*ccx_s_options).scc_framerate = options.scc_framerate;
     // Also copy to enc_cfg so the encoder uses the same frame rate for SCC output
     (*ccx_s_options).enc_cfg.scc_framerate = options.scc_framerate;
+    (*ccx_s_options).enc_cfg.scc_accurate_timing = options.scc_accurate_timing.into();
     #[cfg(feature = "with_libcurl")]
     {
         if options.curlposturl.is_some() {
@@ -419,13 +416,13 @@ pub unsafe fn copy_to_rust(ccx_s_options: *const ccx_s_options) -> Options {
     options.ocr_oem = (*ccx_s_options).ocr_oem as i8;
     options.psm = (*ccx_s_options).psm;
     options.ocr_quantmode = (*ccx_s_options).ocr_quantmode as u8;
+    options.ocr_line_split = (*ccx_s_options).ocr_line_split != 0;
+    options.ocr_blacklist = (*ccx_s_options).ocr_blacklist != 0;
 
-    // Handle mkvlang (C string to Option<Language>)
+    // Handle mkvlang (C string to Option<MkvLangFilter>)
     if !(*ccx_s_options).mkvlang.is_null() {
-        options.mkvlang = Some(
-            Language::from_str(&c_char_to_string((*ccx_s_options).mkvlang))
-                .expect("Invalid language"),
-        )
+        let lang_str = c_char_to_string((*ccx_s_options).mkvlang);
+        options.mkvlang = MkvLangFilter::new(&lang_str).ok();
     }
 
     options.analyze_video_stream = (*ccx_s_options).analyze_video_stream != 0;
@@ -535,6 +532,7 @@ pub unsafe fn copy_to_rust(ccx_s_options: *const ccx_s_options) -> Options {
     options.out_interval = (*ccx_s_options).out_interval;
     options.segment_on_key_frames_only = (*ccx_s_options).segment_on_key_frames_only != 0;
     options.scc_framerate = (*ccx_s_options).scc_framerate;
+    options.scc_accurate_timing = (*ccx_s_options).enc_cfg.scc_accurate_timing != 0;
 
     // Handle optional features with conditional compilation
     #[cfg(feature = "with_libcurl")]
@@ -978,6 +976,7 @@ impl CType<encoder_cfg> for EncoderConfig {
             },
             extract_only_708: self.extract_only_708 as _,
             scc_framerate: 0, // Will be set from ccx_options.scc_framerate in copy_to_c
+            scc_accurate_timing: 0, // Will be set from ccx_options.scc_accurate_timing in copy_to_c
         }
     }
 }
@@ -1070,7 +1069,6 @@ impl CType<program_info> for ProgramInfo {
         program_info {
             pid: self.pid,
             program_number: self.program_number,
-            initialized_ocr: self.initialized_ocr as c_int,
             _bitfield_align_1: [],
             _bitfield_1: bf1,
             version: self.version,

src/rust/src/ctorust.rs

@@ -540,7 +540,6 @@ impl FromCType<program_info> for ProgramInfo {
         Some(ProgramInfo {
             pid: info.pid,
             program_number: info.program_number,
-            initialized_ocr: info.initialized_ocr != 0,
             analysed_pmt_once: info._bitfield_1.get_bit(0) as u8,
             version: info.version,
             saved_section: info.saved_section,
@@ -616,50 +615,6 @@ impl FromCType<ccx_demux_report> for CcxDemuxReport {
     }
 }
 
-/// # Safety
-/// This function is unsafe because it takes a raw pointer to a C struct.
-impl FromCType<*mut PMT_entry> for *mut PMTEntry {
-    unsafe fn from_ctype(buffer_ptr: *mut PMT_entry) -> Option<Self> {
-        if buffer_ptr.is_null() {
-            return None;
-        }
-
-        let buffer = unsafe { &*buffer_ptr };
-
-        let program_number = if buffer.program_number != 0 {
-            buffer.program_number
-        } else {
-            0
-        };
-
-        let elementary_pid = if buffer.elementary_PID != 0 {
-            buffer.elementary_PID
-        } else {
-            0
-        };
-
-        let stream_type = if buffer.stream_type != 0 {
-            StreamType::from_ctype(buffer.stream_type as u32).unwrap_or(StreamType::Unknownstream)
-        } else {
-            StreamType::Unknownstream
-        };
-
-        let printable_stream_type = if buffer.printable_stream_type != 0 {
-            buffer.printable_stream_type
-        } else {
-            0
-        };
-
-        let mut pmt_entry = PMTEntry {
-            program_number,
-            elementary_pid,
-            stream_type,
-            printable_stream_type,
-        };
-
-        Some(&mut pmt_entry as *mut PMTEntry)
-    }
-}
 impl FromCType<ccx_bufferdata_type> for BufferdataType {
     unsafe fn from_ctype(c_value: ccx_bufferdata_type) -> Option<Self> {
         let rust_value = match c_value {

src/rust/src/decoder/mod.rs

@@ -10,7 +10,10 @@ mod timing;
 mod tv_screen;
 mod window;
 
+use log::debug as log_debug;
+
 use lib_ccxr::{
+    common::DTVCC_MAX_SERVICES,
     debug, fatal,
     util::log::{DebugMessageFlag, ExitCause},
 };
@@ -208,6 +211,361 @@ impl<'a> Dtvcc<'a> {
     }
 }
 
+// =============================================================================
+// DtvccRust: Persistent CEA-708 decoder context for Rust-owned state
+// =============================================================================
+//
+// This struct is designed to be created once and persist throughout the program's
+// lifetime, solving the issue where state was being reset on each call.
+// See: https://github.com/CCExtractor/ccextractor/issues/1499
+
+/// Persistent CEA-708 decoder context that owns its data.
+///
+/// Unlike `Dtvcc` which borrows from C structures, `DtvccRust` owns all its
+/// decoder state and is designed to persist across multiple processing calls.
+/// This is created once via `ccxr_dtvcc_init()` and freed via `ccxr_dtvcc_free()`.
+pub struct DtvccRust {
+    pub is_active: bool,
+    pub active_services_count: u8,
+    pub services_active: Vec<i32>,
+    pub report_enabled: bool,
+    pub report: *mut ccx_decoder_dtvcc_report,
+    pub decoders: [Option<Box<dtvcc_service_decoder>>; DTVCC_MAX_SERVICES],
+    pub packet: Vec<u8>,
+    pub packet_length: u8,
+    pub is_header_parsed: bool,
+    pub last_sequence: i32,
+    pub encoder: *mut encoder_ctx,
+    pub no_rollup: bool,
+    pub timing: *mut ccx_common_timing_ctx,
+}
+
+impl DtvccRust {
+    /// Create a new persistent dtvcc context from settings.
+    ///
+    /// This closely follows `dtvcc_init` at `src/lib_ccx/ccx_dtvcc.c:82`
+    ///
+    /// # Safety
+    /// The following pointers in `opts` must not be null:
+    /// - `opts.report`
+    /// - `opts.timing`
+    pub fn new(opts: &ccx_decoder_dtvcc_settings) -> Self {
+        let is_active = is_true(opts.enabled);
+        let active_services_count = opts.active_services_count as u8;
+        let services_active = opts.services_enabled.to_vec();
+        let report_enabled = is_true(opts.print_file_reports);
+
+        // Reset the report counter
+        if !opts.report.is_null() {
+            unsafe {
+                (*opts.report).reset_count = 0;
+            }
+        }
+
+        // Initialize packet state (equivalent to dtvcc_clear_packet)
+        let packet_length = 0;
+        let is_header_parsed = false;
+        let packet = vec![0u8; CCX_DTVCC_MAX_PACKET_LENGTH as usize];
+
+        let last_sequence = CCX_DTVCC_NO_LAST_SEQUENCE;
+        let no_rollup = is_true(opts.no_rollup);
+
+        // Initialize decoders - only for active services
+        // Note: dtvcc_service_decoder is a large struct, so we must allocate it
+        // directly on the heap to avoid stack overflow.
+        let decoders = {
+            const INIT: Option<Box<dtvcc_service_decoder>> = None;
+            let mut decoders = [INIT; DTVCC_MAX_SERVICES];
+
+            for (i, d) in decoders.iter_mut().enumerate() {
+                if i >= opts.services_enabled.len() || !is_true(opts.services_enabled[i]) {
+                    continue;
+                }
+
+                // Create owned tv_screen on the heap using zeroed allocation
+                // to avoid stack overflow (dtvcc_tv_screen is also large)
+                let tv_layout = std::alloc::Layout::new::<dtvcc_tv_screen>();
+                let tv_ptr = unsafe { std::alloc::alloc_zeroed(tv_layout) } as *mut dtvcc_tv_screen;
+                if tv_ptr.is_null() {
+                    panic!("Failed to allocate dtvcc_tv_screen");
+                }
+                let mut tv_screen = unsafe { Box::from_raw(tv_ptr) };
+                tv_screen.cc_count = 0;
+                tv_screen.service_number = i as i32 + 1;
+
+                // Allocate decoder directly on heap using zeroed memory to avoid
+                // stack overflow (dtvcc_service_decoder is very large)
+                let decoder_layout = std::alloc::Layout::new::<dtvcc_service_decoder>();
+                let decoder_ptr = unsafe { std::alloc::alloc_zeroed(decoder_layout) }
+                    as *mut dtvcc_service_decoder;
+                if decoder_ptr.is_null() {
+                    panic!("Failed to allocate dtvcc_service_decoder");
+                }
+
+                let mut decoder = unsafe { Box::from_raw(decoder_ptr) };
+
+                // Set the tv pointer
+                decoder.tv = Box::into_raw(tv_screen);
+
+                // Initialize windows
+                for window in decoder.windows.iter_mut() {
+                    window.memory_reserved = 0;
+                }
+
+                // Call reset handler
+                decoder.handle_reset();
+
+                *d = Some(decoder);
+            }
+
+            decoders
+        };
+
+        // Encoder is set later via set_encoder()
+        let encoder = std::ptr::null_mut();
+
+        DtvccRust {
+            is_active,
+            active_services_count,
+            services_active,
+            report_enabled,
+            report: opts.report,
+            decoders,
+            packet,
+            packet_length,
+            is_header_parsed,
+            last_sequence,
+            no_rollup,
+            timing: opts.timing,
+            encoder,
+        }
+    }
+
+    /// Set the encoder for this context.
+    ///
+    /// The encoder is typically not available at initialization time,
+    /// so it must be set separately before processing.
+    pub fn set_encoder(&mut self, encoder: *mut encoder_ctx) {
+        self.encoder = encoder;
+    }
+
+    /// Process cc data and add it to the dtvcc packet.
+    ///
+    /// This is the main entry point for CEA-708 data processing.
+    pub fn process_cc_data(&mut self, cc_valid: u8, cc_type: u8, data1: u8, data2: u8) {
+        if !self.is_active && !self.report_enabled {
+            return;
+        }
+
+        match cc_type {
+            // type 0 and 1 are for CEA 608 data and are handled before calling this function
+            // valid types for CEA 708 data are only 2 and 3
+            2 => {
+                log_debug!("dtvcc_process_data: DTVCC Channel Packet Data");
+                if cc_valid == 1 && self.is_header_parsed {
+                    if self.packet_length > 253 {
+                        log_debug!("dtvcc_process_data: Warning: Legal packet size exceeded (1), data not added.");
+                    } else {
+                        self.add_data_to_packet(data1, data2);
+
+                        let mut max_len = self.packet[0] & 0x3F;
+
+                        if max_len == 0 {
+                            // This is well defined in EIA-708; no magic.
+                            max_len = 128;
+                        } else {
+                            max_len *= 2;
+                        }
+
+                        // If packet is complete then process the packet
+                        if self.packet_length >= max_len {
+                            self.process_current_packet(max_len);
+                        }
+                    }
+                }
+            }
+            3 => {
+                log_debug!("dtvcc_process_data: DTVCC Channel Packet Start");
+                if cc_valid == 1 {
+                    if self.packet_length > (CCX_DTVCC_MAX_PACKET_LENGTH - 1) {
+                        log_debug!("dtvcc_process_data: Warning: Legal packet size exceeded (2), data not added.");
+                    } else {
+                        if self.is_header_parsed {
+                            log_debug!("dtvcc_process_data: Warning: Incorrect packet length specified. Packet will be skipped.");
+                            self.clear_packet();
+                        }
+                        self.add_data_to_packet(data1, data2);
+                        self.is_header_parsed = true;
+                    }
+                }
+            }
+            _ => fatal!(cause = ExitCause::Bug;
+                "dtvcc_process_data: shouldn't be here - cc_type: {}",
+                cc_type
+            ),
+        }
+    }
+
+    /// Add data to the packet
+    fn add_data_to_packet(&mut self, data1: u8, data2: u8) {
+        self.packet[self.packet_length as usize] = data1;
+        self.packet_length += 1;
+        self.packet[self.packet_length as usize] = data2;
+        self.packet_length += 1;
+    }
+
+    /// Process current packet into service blocks
+    fn process_current_packet(&mut self, len: u8) {
+        let seq = (self.packet[0] & 0xC0) >> 6;
+        log_debug!("dtvcc_process_current_packet: Sequence: {seq}, packet length: {len}");
+        if self.packet_length == 0 {
+            return;
+        }
+
+        // Check if current sequence is correct
+        // Sequence number is a 2 bit rolling sequence from (0-3)
+        if self.last_sequence != CCX_DTVCC_NO_LAST_SEQUENCE
+            && (self.last_sequence + 1) % 4 != seq as i32
+        {
+            log_debug!(
+                "dtvcc_process_current_packet: Unexpected sequence number, it is {} but should be {}",
+                seq, (self.last_sequence + 1) % 4
+            );
+        }
+        self.last_sequence = seq as i32;
+
+        let mut pos: u8 = 1;
+        while pos < len {
+            let mut service_number = (self.packet[pos as usize] & 0xE0) >> 5; // 3 more significant bits
+            let block_length = self.packet[pos as usize] & 0x1F; // 5 less significant bits
+            log_debug!("dtvcc_process_current_packet: Standard header Service number: {service_number}, Block length: {block_length}");
+
+            if service_number == 7 {
+                // There is an extended header
+                // CEA-708-E 6.2.2 Extended Service Block Header
+                pos += 1;
+                service_number = self.packet[pos as usize] & 0x3F; // 6 more significant bits
+                if service_number > 7 {
+                    log_debug!("dtvcc_process_current_packet: Illegal service number in extended header: {service_number}");
+                }
+            }
+
+            pos += 1;
+
+            if service_number == 0 && block_length != 0 {
+                // Illegal, but specs say what to do...
+                pos = len; // Move to end
+                break;
+            }
+
+            if block_length != 0 && !self.report.is_null() {
+                unsafe {
+                    (*self.report).services[service_number as usize] = 1;
+                }
+            }
+
+            if service_number > 0 && is_true(self.services_active[(service_number - 1) as usize]) {
+                if let Some(decoder) = &mut self.decoders[(service_number - 1) as usize] {
+                    // Get encoder and timing references
+                    if !self.encoder.is_null() && !self.timing.is_null() {
+                        let encoder = unsafe { &mut *self.encoder };
+                        let timing = unsafe { &mut *self.timing };
+                        decoder.process_service_block(
+                            &self.packet[pos as usize..(pos + block_length) as usize],
+                            encoder,
+                            timing,
+                            self.no_rollup,
+                        );
+                    }
+                }
+            }
+
+            pos += block_length // Skip data
+        }
+
+        self.clear_packet();
+
+        if len < 128 && self.packet[pos as usize] != 0 {
+            // Null header is mandatory if there is room
+            log_debug!(
+                "dtvcc_process_current_packet: Warning: Null header expected but not found."
+            );
+        }
+    }
+
+    /// Clear current packet
+    fn clear_packet(&mut self) {
+        self.packet_length = 0;
+        self.is_header_parsed = false;
+        self.packet.iter_mut().for_each(|x| *x = 0);
+    }
+
+    /// Flush all active service decoders.
+    ///
+    /// This writes out any pending caption data from all active services.
+    /// Called when processing is complete or when switching contexts.
+    pub fn flush_active_decoders(&mut self) {
+        if !self.is_active {
+            return;
+        }
+
+        for i in 0..DTVCC_MAX_SERVICES {
+            if i >= self.services_active.len() || !is_true(self.services_active[i]) {
+                continue;
+            }
+
+            if let Some(decoder) = &mut self.decoders[i] {
+                // Check if there's content to flush: either cc_count > 0 (already printed)
+                // or any window has visible content (needs to be printed during flush)
+                let has_visible_windows = decoder.windows.iter().any(|w| is_true(w.visible));
+                if decoder.cc_count > 0 || has_visible_windows {
+                    self.flush_decoder(i);
+                }
+            }
+        }
+    }
+
+    /// Flush a specific service decoder by index.
+    fn flush_decoder(&mut self, service_index: usize) {
+        log_debug!(
+            "dtvcc_decoder_flush: Flushing decoder for service {}",
+            service_index + 1
+        );
+
+        // Need encoder and timing to flush
+        if self.encoder.is_null() || self.timing.is_null() {
+            log_debug!("dtvcc_decoder_flush: Cannot flush - encoder or timing is null");
+            return;
+        }
+
+        if let Some(decoder) = &mut self.decoders[service_index] {
+            let timing = unsafe { &mut *self.timing };
+            let encoder = unsafe { &mut *self.encoder };
+
+            let mut screen_content_changed = false;
+
+            // Process all visible windows
+            for i in 0..CCX_DTVCC_MAX_WINDOWS {
+                let window = &mut decoder.windows[i as usize];
+                if is_true(window.visible) {
+                    screen_content_changed = true;
+                    window.update_time_hide(timing);
+                    // Copy window content to screen
+                    decoder.copy_to_screen(&decoder.windows[i as usize]);
+                    decoder.windows[i as usize].visible = 0;
+                }
+            }
+
+            if screen_content_changed {
+                decoder.screen_print(encoder, timing);
+            }
+            decoder.flush(encoder);
+        }
+    }
+}
+
+const CCX_DTVCC_MAX_WINDOWS: u8 = 8;
+
 /// A single character symbol
 ///
 /// sym stores the symbol
@@ -361,4 +719,130 @@ pub mod test {
         assert_eq!(decoder.report.services[8], 1);
         assert_eq!(decoder.packet_length, 0); // due to `clear_packet()` fn call
     }
+
+    // =========================================================================
+    // Tests for DtvccRust (persistent CEA-708 decoder)
+    // =========================================================================
+
+    /// Helper function to create a test ccx_decoder_dtvcc_settings
+    /// Uses heap allocation to avoid stack overflow with large structs
+    pub fn create_test_dtvcc_settings() -> Box<ccx_decoder_dtvcc_settings> {
+        let mut settings = get_zero_allocated_obj::<ccx_decoder_dtvcc_settings>();
+
+        // Initialize required pointers using heap allocation
+        let report = get_zero_allocated_obj::<ccx_decoder_dtvcc_report>();
+        settings.report = Box::into_raw(report);
+
+        let timing = get_zero_allocated_obj::<ccx_common_timing_ctx>();
+        settings.timing = Box::into_raw(timing);
+
+        // Enable the decoder and first service
+        settings.enabled = 1;
+        settings.active_services_count = 1;
+        settings.services_enabled[0] = 1;
+
+        settings
+    }
+
+    #[test]
+    fn test_dtvcc_rust_new() {
+        let settings = create_test_dtvcc_settings();
+        let dtvcc = DtvccRust::new(&settings);
+
+        // Verify basic initialization
+        assert!(dtvcc.is_active);
+        assert_eq!(dtvcc.active_services_count, 1);
+        assert_eq!(dtvcc.packet_length, 0);
+        assert!(!dtvcc.is_header_parsed);
+        assert_eq!(dtvcc.last_sequence, CCX_DTVCC_NO_LAST_SEQUENCE);
+
+        // Verify encoder is initially null (set later)
+        assert!(dtvcc.encoder.is_null());
+
+        // Verify first decoder is created (service 0 is enabled)
+        assert!(dtvcc.decoders[0].is_some());
+
+        // Verify other decoders are not created
+        assert!(dtvcc.decoders[1].is_none());
+    }
+
+    #[test]
+    fn test_dtvcc_rust_set_encoder() {
+        let settings = create_test_dtvcc_settings();
+        let mut dtvcc = DtvccRust::new(&settings);
+
+        // Initially null
+        assert!(dtvcc.encoder.is_null());
+
+        // Create an encoder and set it
+        let mut encoder = Box::new(encoder_ctx::default());
+        let encoder_ptr = &mut *encoder as *mut encoder_ctx;
+        dtvcc.set_encoder(encoder_ptr);
+
+        // Verify encoder is set
+        assert!(!dtvcc.encoder.is_null());
+        assert_eq!(dtvcc.encoder, encoder_ptr);
+    }
+
+    #[test]
+    fn test_dtvcc_rust_process_cc_data() {
+        let settings = create_test_dtvcc_settings();
+        let mut dtvcc = DtvccRust::new(&settings);
+
+        // Process cc_type = 3 (packet start) - should set is_header_parsed
+        dtvcc.process_cc_data(1, 3, 0xC2, 0x00);
+
+        assert!(dtvcc.is_header_parsed);
+        assert_eq!(dtvcc.packet_length, 2);
+        assert_eq!(dtvcc.packet[0], 0xC2);
+        assert_eq!(dtvcc.packet[1], 0x00);
+    }
+
+    #[test]
+    fn test_dtvcc_rust_clear_packet() {
+        let settings = create_test_dtvcc_settings();
+        let mut dtvcc = DtvccRust::new(&settings);
+
+        // Add some data
+        dtvcc.process_cc_data(1, 3, 0xC2, 0x00);
+        assert!(dtvcc.is_header_parsed);
+        assert_eq!(dtvcc.packet_length, 2);
+
+        // Process more data that triggers clear (when packet is malformed)
+        // Simulate by directly testing the packet processing
+        dtvcc.is_header_parsed = true;
+        dtvcc.packet[0] = 0x02; // Very short packet length (2*1 = 2 bytes)
+        dtvcc.packet_length = 2;
+
+        // This should process and clear the packet
+        dtvcc.process_cc_data(1, 2, 0x00, 0x00);
+
+        // After processing a complete packet, it should be cleared
+        assert_eq!(dtvcc.packet_length, 0);
+        assert!(!dtvcc.is_header_parsed);
+    }
+
+    #[test]
+    fn test_dtvcc_rust_state_persistence() {
+        // This test verifies the key fix: state persists across calls
+        let settings = create_test_dtvcc_settings();
+        let mut dtvcc = DtvccRust::new(&settings);
+
+        // First call: start a packet
+        dtvcc.process_cc_data(1, 3, 0xC4, 0x00); // Packet with length 4*2=8 bytes
+        assert!(dtvcc.is_header_parsed);
+        assert_eq!(dtvcc.packet_length, 2);
+
+        // Second call: add more data (this is where the old code would fail)
+        dtvcc.process_cc_data(1, 2, 0x21, 0x00);
+        assert_eq!(dtvcc.packet_length, 4);
+
+        // Third call: add more data
+        dtvcc.process_cc_data(1, 2, 0x00, 0x00);
+        assert_eq!(dtvcc.packet_length, 6);
+
+        // State is preserved across all calls!
+        assert!(dtvcc.is_header_parsed);
+        assert_eq!(dtvcc.last_sequence, CCX_DTVCC_NO_LAST_SEQUENCE); // Not processed yet
+    }
 }

src/rust/src/decoder/service_decoder.rs

@@ -1259,6 +1259,7 @@ extern "C" fn ccxr_flush_decoder(dtvcc: *mut dtvcc_ctx, decoder: *mut dtvcc_serv
 mod test {
     use super::*;
     use crate::utils::get_zero_allocated_obj;
+    use std::alloc::{alloc_zeroed, dealloc, Layout};
 
     fn setup_test_decoder_with_memory() -> dtvcc_service_decoder {
         let mut decoder = get_zero_allocated_obj::<dtvcc_service_decoder>();
@@ -1349,10 +1350,17 @@ mod test {
         decoder.current_window = 1;
         decoder.windows[1].pen_column = 12;
         decoder.windows[1].pen_row = 1;
-        decoder.windows[1].rows[1] = Box::into_raw(Box::new(dtvcc_symbol::new(1)));
-        decoder.windows[1].rows[2] = Box::into_raw(Box::new(dtvcc_symbol::new(1)));
+        let layout = Layout::array::<dtvcc_symbol>(CCX_DTVCC_MAX_COLUMNS as usize).unwrap();
+        for i in 0..CCX_DTVCC_MAX_ROWS as usize {
+            decoder.windows[1].rows[i] = unsafe { alloc_zeroed(layout) } as *mut dtvcc_symbol;
+        }
         decoder.windows[1].memory_reserved = 1;
 
+        unsafe {
+            *decoder.windows[1].rows[1] = dtvcc_symbol::new(1);
+            *decoder.windows[1].rows[2] = dtvcc_symbol::new(1);
+        }
+
         decoder.process_hcr();
 
         assert_eq!(decoder.windows[1].pen_column, 0);
@@ -1367,6 +1375,13 @@ mod test {
             unsafe { decoder.windows[1].rows[2].as_mut() },
             Some(&mut dtvcc_symbol { sym: 1, init: 1 }),
         );
+
+        // Cleanup
+        for i in 0..CCX_DTVCC_MAX_ROWS as usize {
+            unsafe {
+                dealloc(decoder.windows[1].rows[i] as *mut u8, layout);
+            }
+        }
     }
 
     #[test]
@@ -1376,8 +1391,16 @@ mod test {
         decoder.windows[1].pen_column = 2;
         decoder.windows[1].pen_row = 1;
         decoder.windows[1].memory_reserved = 1;
-        decoder.windows[1].rows[1] = Box::into_raw(Box::new(dtvcc_symbol::new(1)));
-        decoder.windows[1].rows[2] = Box::into_raw(Box::new(dtvcc_symbol::new(1)));
+        let layout = Layout::array::<dtvcc_symbol>(CCX_DTVCC_MAX_COLUMNS as usize).unwrap();
+        for i in 0..CCX_DTVCC_MAX_ROWS as usize {
+            decoder.windows[1].rows[i] = unsafe { alloc_zeroed(layout) } as *mut dtvcc_symbol;
+        }
+        decoder.windows[1].memory_reserved = 1;
+
+        unsafe {
+            *decoder.windows[1].rows[1] = dtvcc_symbol::new(1);
+            *decoder.windows[1].rows[2] = dtvcc_symbol::new(1);
+        }
 
         decoder.process_ff();
 
@@ -1394,6 +1417,13 @@ mod test {
             unsafe { decoder.windows[1].rows[2].as_mut() },
             Some(&mut dtvcc_symbol::default()),
         );
+
+        // Cleanup
+        for i in 0..CCX_DTVCC_MAX_ROWS as usize {
+            unsafe {
+                dealloc(decoder.windows[1].rows[i] as *mut u8, layout);
+            }
+        }
     }
 
     #[test]

src/rust/src/decoder/window.rs

@@ -167,7 +167,9 @@ impl dtvcc_window {
                 } else {
                     let layout = layout.unwrap();
                     // deallocate previous memory
-                    dealloc(self.rows[row_index] as *mut u8, layout);
+                    if !self.rows[row_index].is_null() {
+                        dealloc(self.rows[row_index] as *mut u8, layout);
+                    }
 
                     // allocate new zero initialized memory
                     let ptr = alloc_zeroed(layout);

src/rust/src/demuxer/common_types.rs

@@ -1,8 +1,13 @@
 use crate::bindings::{lib_ccx_ctx, list_head};
 use lib_ccxr::common::{Codec, Decoder608Report, DecoderDtvccReport, StreamMode, StreamType};
 use lib_ccxr::time::Timestamp;
+use std::os::raw::c_void;
 use std::ptr::null_mut;
 
+extern "C" {
+    fn free(ptr: *mut c_void);
+}
+
 // Size of the Startbytes Array in CcxDemuxer - const 1MB
 pub(crate) const ARRAY_SIZE: usize = 1024 * 1024;
 
@@ -50,7 +55,6 @@ pub struct FileReport {
 pub struct ProgramInfo {
     pub pid: i32,
     pub program_number: i32,
-    pub initialized_ocr: bool, // Avoid initializing the OCR more than once
     pub analysed_pmt_once: u8, // 1-bit field
     pub version: u8,
     pub saved_section: [u8; SAVED_SECTIONS_PROGRAMINFO],
@@ -110,7 +114,9 @@ impl Default for PSIBuffer {
     fn default() -> Self {
         PSIBuffer {
             prev_ccounter: 0,
-            buffer: Box::into_raw(Box::new(0u8)),
+            // Initialize with null to avoid unnecessary heap allocations and
+            // signal that the buffer is currently empty.
+            buffer: std::ptr::null_mut(),
             buffer_length: 0,
             ccounter: 0,
         }
@@ -275,21 +281,21 @@ impl Default for CcxDemuxer<'_> {
 /// null pointers which are safely ignored.
 impl Drop for CcxDemuxer<'_> {
     fn drop(&mut self) {
-        // Free all non-null PSIBuffer pointers (Rust-owned from Box::into_raw)
+        // Free all non-null PSIBuffer pointers.
+        // These are freed using C's free to be compatible with memory that might be allocated by C.
         for ptr in self.pid_buffers.drain(..) {
             if !ptr.is_null() {
-                // SAFETY: These pointers were created via Box::into_raw in copy_demuxer_from_c_to_rust
                 unsafe {
-                    drop(Box::from_raw(ptr));
+                    free(ptr as *mut c_void);
                 }
             }
         }
-        // Free all non-null PMTEntry pointers (Rust-owned from Box::into_raw)
+        // Free all non-null PMTEntry pointers.
+        // These are freed using C's free to be compatible with memory that might be allocated by C.
         for ptr in self.pids_programs.drain(..) {
             if !ptr.is_null() {
-                // SAFETY: These pointers were created via Box::into_raw in copy_demuxer_from_c_to_rust
                 unsafe {
-                    drop(Box::from_raw(ptr));
+                    free(ptr as *mut c_void);
                 }
             }
         }
@@ -301,7 +307,6 @@ impl Default for ProgramInfo {
         ProgramInfo {
             pid: -1,
             program_number: 0,
-            initialized_ocr: false,
             analysed_pmt_once: 0,
             version: 0,
             saved_section: [0; SAVED_SECTIONS_PROGRAMINFO],

src/rust/src/demuxer/demux.rs

@@ -327,6 +327,9 @@ impl CcxDemuxer<'_> {
             StreamMode::Mxf => {
                 info!("MXF");
             }
+            StreamMode::Scc => {
+                info!("SCC");
+            }
             #[cfg(feature = "wtv_debug")]
             StreamMode::HexDump => {
                 info!("Hex");

src/rust/src/demuxer/stream_functions.rs

@@ -331,10 +331,15 @@ unsafe fn detect_stream_type_common(ctx: &mut CcxDemuxer, ccx_options: &mut Opti
             }
 
             // Now check for PS (Needs PACK header)
+            // The loop below checks 4 consecutive bytes (i, i+1, i+2, i+3), so we need
+            // to stop 3 bytes before the end to avoid out-of-bounds access.
+            // - If buffer < 50000: limit = buffer_size - 3 (scan entire buffer)
+            // - If buffer >= 50000: limit = 49997 (= 50000 - 3, cap the scan range)
+            // We use saturating_sub to safely handle tiny buffers (< 3 bytes).
             let limit = if ctx.startbytes_avail < 50000 {
-                ctx.startbytes_avail - 3
+                ctx.startbytes_avail.saturating_sub(3)
             } else {
-                49997
+                50000 - 3 // Don't scan huge buffers entirely; 50KB is enough
             } as usize;
             for i in 0..limit {
                 if ctx.startbytes[i] == 0x00
@@ -427,15 +432,21 @@ pub fn is_valid_mp4_box(
                 )
             );
 
-            // If the box type is "moov", check if it contains a valid movie header (mvhd)
-            if idx == 2
-                && !(buffer[position + 12] == b'm'
+            // If the box type is "moov", it must contain "mvhd" to be valid.
+            // We need 16 bytes from position to check bytes 12-15 for "mvhd".
+            if idx == 2 {
+                if position + 16 > buffer.len() {
+                    // Not enough bytes to verify mvhd - skip this box
+                    continue;
+                }
+                if !(buffer[position + 12] == b'm'
                     && buffer[position + 13] == b'v'
                     && buffer[position + 14] == b'h'
                     && buffer[position + 15] == b'd')
-            {
-                // If "moov" doesn't have "mvhd", skip it.
-                continue;
+                {
+                    // moov without mvhd is not valid - skip it
+                    continue;
+                }
             }
 
             // Box name matches. Do a crude validation of possible box size,

src/rust/src/es/userdata.rs

@@ -278,7 +278,8 @@ pub unsafe fn user_data(
 
                 if !proceed {
                     debug!(msg_type = DebugMessageFlag::VERBOSE; "\rThe following payload is not properly terminated.");
-                    dump(cc_data.to_vec().as_mut_ptr(), (cc_count * 3 + 1) as _, 0, 0);
+                    let mut cc_data_copy = cc_data.to_vec();
+                    dump(cc_data_copy.as_mut_ptr(), (cc_count * 3 + 1) as _, 0, 0);
                 }
                 debug!(msg_type = DebugMessageFlag::VERBOSE; "Reading {} HD CC blocks", cc_count);
 
@@ -289,10 +290,11 @@ pub unsafe fn user_data(
                 // Please note we store the current value of the global
                 // fts_now variable (and not get_fts()) as we are going to
                 // re-create the timeline in process_hdcc() (Slightly ugly).
+                let mut cc_data_copy = cc_data.to_vec();
                 store_hdcc(
                     enc_ctx,
                     dec_ctx,
-                    cc_data.to_vec().as_mut_ptr(),
+                    cc_data_copy.as_mut_ptr(),
                     cc_count as _,
                     (*dec_ctx.timing).current_tref,
                     (*dec_ctx.timing).fts_now,
@@ -340,6 +342,10 @@ pub unsafe fn user_data(
         let dcd_pos = ustream.pos; // dish caption data position
         match pattern_type {
             0x02 => {
+                if ustream.data.len() - ustream.pos < 4 {
+                    info!("Dish Network caption: insufficient data");
+                    return Ok(1);
+                }
                 // Two byte caption - always on B-frame
                 // The following 4 bytes are:
                 // 0  :  0x09
@@ -387,6 +393,10 @@ pub unsafe fn user_data(
                 // Ignore 3 (0x0A, followed by two unknown) bytes.
             }
             0x04 => {
+                if ustream.data.len() - ustream.pos < 5 {
+                    info!("Dish Network caption: insufficient data");
+                    return Ok(1);
+                }
                 // Four byte caption - always on B-frame
                 // The following 5 bytes are:
                 // 0  :  0x09
@@ -423,6 +433,10 @@ pub unsafe fn user_data(
                 // Ignore 4 (0x020A, followed by two unknown) bytes.
             }
             0x05 => {
+                if ustream.data.len() - ustream.pos < 12 {
+                    info!("Dish Network caption: insufficient data");
+                    return Ok(1);
+                }
                 // Buffered caption - always on I-/P-frame
                 // The following six bytes are:
                 // 0  :  0x04
@@ -430,7 +444,7 @@ pub unsafe fn user_data(
                 // 1  : prev dcd[2]
                 // 2-3: prev dcd[3-4]
                 // 4-5: prev dcd[5-6]
-                let dcd_data = &ustream.data[dcd_pos..dcd_pos + 10]; // Need more bytes for this case
+                let dcd_data = &ustream.data[dcd_pos..dcd_pos + 12]; // Need more bytes for this case
                 debug!(msg_type = DebugMessageFlag::PARSE; " - {:02X}  pch: {:02X} {:5} {:02X}:{:02X}",
                           dcd_data[0], dcd_data[1],
                           (dcd_data[2] as u32) * 256 + (dcd_data[3] as u32),
@@ -532,10 +546,12 @@ pub unsafe fn user_data(
 
         if udatalen < 720 {
             info!("MPEG:VBI: Minimum 720 bytes in luma line required");
+            return Ok(1);
         }
 
         let vbi_data = &ustream.data[ustream.pos..ustream.pos + 720];
-        decode_vbi(dec_ctx, field, vbi_data.to_vec().as_mut_ptr(), 720, sub);
+        let mut vbi_data_copy = vbi_data.to_vec();
+        decode_vbi(dec_ctx, field, vbi_data_copy.as_mut_ptr(), 720, sub);
         debug!(msg_type = DebugMessageFlag::VERBOSE; "GXF (vbi line {}) user data:", line_nb);
     } else {
         // Some other user data
@@ -543,14 +559,8 @@ pub unsafe fn user_data(
         debug!(msg_type = DebugMessageFlag::VERBOSE; "Unrecognized user data:");
         let udatalen = ustream.data.len() - ustream.pos;
         let dump_len = if udatalen > 128 { 128 } else { udatalen };
-        dump(
-            ustream.data[ustream.pos..ustream.pos + dump_len]
-                .to_vec()
-                .as_mut_ptr(),
-            dump_len as _,
-            0,
-            0,
-        );
+        let mut data_copy = ustream.data[ustream.pos..ustream.pos + dump_len].to_vec();
+        dump(data_copy.as_mut_ptr(), dump_len as _, 0, 0);
     }
 
     debug!(msg_type = DebugMessageFlag::VERBOSE; "User data - processed");

src/rust/src/file_functions/file.rs

@@ -129,10 +129,14 @@ pub fn sleepandchecktimeout(start: u64, ccx_options: &mut Options) {
         .expect("System time went backwards")
         .as_secs();
 
-    if ccx_options.live_stream.is_some() && ccx_options.live_stream.unwrap().seconds() != 0 {
-        if current_time > start + ccx_options.live_stream.unwrap().millis() as u64 {
-            // Timeout elapsed
-            ccx_options.live_stream = Option::from(Timestamp::from_millis(0));
+    if let Some(live_stream) = ccx_options.live_stream {
+        if live_stream.seconds() != 0 {
+            if current_time > start + live_stream.millis() as u64 {
+                // Timeout elapsed
+                ccx_options.live_stream = Option::from(Timestamp::from_millis(0));
+            } else {
+                sleep_secs(1);
+            }
         } else {
             sleep_secs(1);
         }

src/rust/src/hardsubx/imgops.rs

@@ -12,7 +12,7 @@ pub extern "C" fn rgb_to_hsv(R: f32, G: f32, B: f32, H: &mut f32, S: &mut f32, V
 
     let hsv_rep = Hsv::from_color(rgb);
 
-    *H = hsv_rep.hue.to_positive_degrees();
+    *H = hsv_rep.hue.into_positive_degrees();
     *S = hsv_rep.saturation;
     *V = hsv_rep.value;
 }

src/rust/src/lib.rs

@@ -38,7 +38,7 @@ use bindings::*;
 use cfg_if::cfg_if;
 use clap::{error::ErrorKind, Parser};
 use common::{copy_from_rust, CType, CType2};
-use decoder::Dtvcc;
+use decoder::{Dtvcc, DtvccRust};
 use lib_ccxr::{common::Options, teletext::TeletextConfig, util::log::ExitCause};
 use parser::OptionsExt;
 use utils::is_true;
@@ -210,11 +210,189 @@ pub extern "C" fn ccxr_init_logger() {
         .init();
 }
 
+// =============================================================================
+// FFI functions for persistent DtvccRust context
+// =============================================================================
+//
+// These functions provide a C-compatible interface for managing the persistent
+// Rust CEA-708 decoder context. They are designed to be called from C code
+// and will be used in Phase 2-3 of the implementation.
+// See: https://github.com/CCExtractor/ccextractor/issues/1499
+
+/// Create a new persistent DtvccRust context.
+///
+/// This function allocates and initializes a new `DtvccRust` struct on the heap
+/// and returns an opaque pointer to it. The context persists until freed with
+/// `ccxr_dtvcc_free()`.
+///
+/// # Safety
+/// - `opts_ptr` must be a valid pointer to `ccx_decoder_dtvcc_settings`
+/// - `opts.report` and `opts.timing` must not be null
+/// - The returned pointer must be freed with `ccxr_dtvcc_free()` when done
+///
+/// # Returns
+/// An opaque pointer to the DtvccRust context, or null if opts_ptr is null.
+#[no_mangle]
+pub unsafe extern "C" fn ccxr_dtvcc_init(
+    opts_ptr: *const ccx_decoder_dtvcc_settings,
+) -> *mut std::ffi::c_void {
+    if opts_ptr.is_null() {
+        return std::ptr::null_mut();
+    }
+    let opts = &*opts_ptr;
+    let dtvcc = Box::new(DtvccRust::new(opts));
+    Box::into_raw(dtvcc) as *mut std::ffi::c_void
+}
+
+/// Free a DtvccRust context.
+///
+/// This function properly frees all memory associated with the DtvccRust context,
+/// including owned decoders and their tv_screens.
+///
+/// # Safety
+/// - `dtvcc_ptr` must be a valid pointer returned by `ccxr_dtvcc_init()`
+/// - `dtvcc_ptr` must not be used after this call
+/// - It is safe to call with a null pointer (no-op)
+#[no_mangle]
+pub extern "C" fn ccxr_dtvcc_free(dtvcc_ptr: *mut std::ffi::c_void) {
+    if dtvcc_ptr.is_null() {
+        return;
+    }
+
+    let dtvcc = unsafe { Box::from_raw(dtvcc_ptr as *mut DtvccRust) };
+
+    // Free owned decoders and their tv_screens
+    for (i, decoder_opt) in dtvcc.decoders.iter().enumerate() {
+        if i >= dtvcc.services_active.len() || !is_true(dtvcc.services_active[i]) {
+            continue;
+        }
+
+        if let Some(decoder) = decoder_opt {
+            // Free windows rows if memory was reserved
+            for window in decoder.windows.iter() {
+                if is_true(window.memory_reserved) {
+                    for row_ptr in window.rows.iter() {
+                        if !row_ptr.is_null() {
+                            unsafe {
+                                drop(Box::from_raw(*row_ptr));
+                            }
+                        }
+                    }
+                }
+            }
+
+            // Free the tv_screen
+            if !decoder.tv.is_null() {
+                unsafe {
+                    drop(Box::from_raw(decoder.tv));
+                }
+            }
+        }
+    }
+
+    // The Box containing dtvcc will be dropped here, freeing the DtvccRust struct
+    drop(dtvcc);
+}
+
+/// Set the encoder for a DtvccRust context.
+///
+/// The encoder is typically not available at initialization time, so it must
+/// be set separately before processing begins.
+///
+/// # Safety
+/// - `dtvcc_ptr` must be a valid pointer returned by `ccxr_dtvcc_init()`
+/// - `encoder` can be null (processing will skip service blocks if so)
+#[no_mangle]
+pub extern "C" fn ccxr_dtvcc_set_encoder(
+    dtvcc_ptr: *mut std::ffi::c_void,
+    encoder: *mut encoder_ctx,
+) {
+    if dtvcc_ptr.is_null() {
+        return;
+    }
+    let dtvcc = unsafe { &mut *(dtvcc_ptr as *mut DtvccRust) };
+    dtvcc.set_encoder(encoder);
+}
+
+/// Process CEA-708 CC data using the persistent DtvccRust context.
+///
+/// This function processes a single CC data unit (cc_valid, cc_type, data1, data2)
+/// using the persistent context, maintaining state across calls.
+///
+/// # Safety
+/// - `dtvcc_ptr` must be a valid pointer returned by `ccxr_dtvcc_init()`
+#[no_mangle]
+pub extern "C" fn ccxr_dtvcc_process_data(
+    dtvcc_ptr: *mut std::ffi::c_void,
+    cc_valid: u8,
+    cc_type: u8,
+    data1: u8,
+    data2: u8,
+) {
+    if dtvcc_ptr.is_null() {
+        return;
+    }
+    let dtvcc = unsafe { &mut *(dtvcc_ptr as *mut DtvccRust) };
+    dtvcc.process_cc_data(cc_valid, cc_type, data1, data2);
+}
+
+/// Flush all active service decoders in the DtvccRust context.
+///
+/// This writes out any pending caption data from all active services.
+/// Should be called when processing is complete or when switching contexts.
+///
+/// # Safety
+/// - `dtvcc_ptr` must be a valid pointer returned by `ccxr_dtvcc_init()`
+/// - It is safe to call with a null pointer (no-op)
+#[no_mangle]
+pub extern "C" fn ccxr_flush_active_decoders(dtvcc_ptr: *mut std::ffi::c_void) {
+    if dtvcc_ptr.is_null() {
+        return;
+    }
+    let dtvcc = unsafe { &mut *(dtvcc_ptr as *mut DtvccRust) };
+    dtvcc.flush_active_decoders();
+}
+
+/// Check if the DtvccRust context is active.
+///
+/// # Safety
+/// - `dtvcc_ptr` must be a valid pointer returned by `ccxr_dtvcc_init()`
+///
+/// # Returns
+/// 1 if active, 0 if not active or if pointer is null.
+#[no_mangle]
+pub extern "C" fn ccxr_dtvcc_is_active(dtvcc_ptr: *mut std::ffi::c_void) -> i32 {
+    if dtvcc_ptr.is_null() {
+        return 0;
+    }
+    let dtvcc = unsafe { &*(dtvcc_ptr as *mut DtvccRust) };
+    if dtvcc.is_active {
+        1
+    } else {
+        0
+    }
+}
+
+/// Enable or disable the DTVCC decoder
+/// This allows enabling the decoder after initialization
+///
+/// # Safety
+/// dtvcc_ptr must be a valid pointer to a DtvccRust struct or null
+#[no_mangle]
+pub extern "C" fn ccxr_dtvcc_set_active(dtvcc_ptr: *mut std::ffi::c_void, active: i32) {
+    if dtvcc_ptr.is_null() {
+        return;
+    }
+    let dtvcc = unsafe { &mut *(dtvcc_ptr as *mut DtvccRust) };
+    dtvcc.is_active = active != 0;
+}
+
 /// Process cc_data
 ///
 /// # Safety
 /// dec_ctx should not be a null pointer
 /// data should point to cc_data of length cc_count
+/// dec_ctx.dtvcc_rust must point to a valid DtvccRust instance
 #[no_mangle]
 extern "C" fn ccxr_process_cc_data(
     dec_ctx: *mut lib_cc_decode,
@@ -228,8 +406,9 @@ extern "C" fn ccxr_process_cc_data(
 
     let dec_ctx = unsafe { &mut *dec_ctx };
 
-    // Check dtvcc pointer before dereferencing
-    if dec_ctx.dtvcc.is_null() {
+    // Check dtvcc_rust pointer before dereferencing (not dtvcc!)
+    // When Rust is enabled, dtvcc is NULL and dtvcc_rust holds the actual context
+    if dec_ctx.dtvcc_rust.is_null() {
         return -1;
     }
 
@@ -237,13 +416,20 @@ extern "C" fn ccxr_process_cc_data(
     let mut cc_data: Vec<u8> = (0..cc_count * 3)
         .map(|x| unsafe { *data.add(x as usize) })
         .collect();
-    let dtvcc_ctx = unsafe { &mut *dec_ctx.dtvcc };
-    let mut dtvcc = Dtvcc::new(dtvcc_ctx);
+
+    // Use the persistent DtvccRust context from dtvcc_rust
+    let dtvcc_rust = dec_ctx.dtvcc_rust as *mut DtvccRust;
+    if dtvcc_rust.is_null() {
+        warn!("ccxr_process_cc_data: dtvcc_rust is null");
+        return ret;
+    }
+    let dtvcc = unsafe { &mut *dtvcc_rust };
+
     for cc_block in cc_data.chunks_exact_mut(3) {
         if !validate_cc_pair(cc_block) {
             continue;
         }
-        let success = do_cb_dtvcc(dec_ctx, &mut dtvcc, cc_block);
+        let success = do_cb_dtvcc_rust(dec_ctx, dtvcc, cc_block);
         if success {
             ret = 0;
         }
@@ -251,11 +437,35 @@ extern "C" fn ccxr_process_cc_data(
     ret
 }
 
-/// Returns `true` if cc_block pair is valid
+/// Validates a closed caption block pair for both CEA-608 and CEA-708 data.
 ///
-/// For CEA-708 data, only cc_valid is checked
-/// For CEA-608 data, parity is also checked
+/// # Arguments
+/// cc_block - A mutable slice containing exactly 3 bytes representing a CC block
+///
+/// # Returns
+/// true if the CC block is valid and should be processed
+/// false if the CC block should be ignored
+///
+/// # Behavior
+/// 1. Header Validation:
+///    - Checks the cc_valid flag (bit 2 of cc_block[0]). If 0, returns false immediately.
+///    - extracts the cc_type (bits 0-1 of cc_block[0]).
+///
+/// 2. CEA-708 (Type 2 or 3):
+///    - No further validation is required beyond the cc_valid flag. Returns true.
+///
+/// 3. CEA-608 (Type 0 or 1):
+///    - Critical Parity Check: Validates parity for the second data byte (cc_block[2]).
+///      If this fails, the entire pair is deemed corrupt, and the function returns false.
+///    - Sanitization: Validates parity for the first data byte (cc_block[1]).
+///      If this fails (but byte 2 was valid), cc_block[1] is overwritten with CC_SOLID_BLANK (0x7F).
+const CC_SOLID_BLANK: u8 = 0x7F;
+
 pub fn validate_cc_pair(cc_block: &mut [u8]) -> bool {
+    if cc_block.len() != 3 {
+        return false;
+    }
+
     let cc_valid = (cc_block[0] & 4) >> 2;
     let cc_type = cc_block[0] & 3;
     if cc_valid == 0 {
@@ -270,7 +480,7 @@ pub fn validate_cc_pair(cc_block: &mut [u8]) -> bool {
         if verify_parity(cc_block[1]) {
             // If the first byte doesn't pass parity,
             // we replace it with a solid blank and process the pair.
-            cc_block[1] = 0x7F;
+            cc_block[1] = CC_SOLID_BLANK;
         }
     }
     true
@@ -280,13 +490,17 @@ pub fn validate_cc_pair(cc_block: &mut [u8]) -> bool {
 ///
 /// CC uses odd parity (i.e., # of 1's in byte is odd.)
 pub fn verify_parity(data: u8) -> bool {
-    if data.count_ones() & 1 == 1 {
-        return true;
-    }
-    false
+    data.count_ones() & 1 == 1
 }
 
-/// Process CC data according to its type
+/// Has different semantic meaning than just a solid blank.
+/// 0x7F can be used as a parity mask to check if the 7 data bits are zero. (0x7F => 0111 1111).
+/// Since the parity bit is forced to be 0 by the mask, it can no longer affect the result.
+/// Therefore, the only way the total result can be 0 is if all the data bits (0-6) were originally zero.
+/// This isn't related to the "solid blank" character - it's just that the mask happens to have the same value.
+const PARITY_BIT_MASK: u8 = 0x7F;
+
+/// Process CC data according to its type (using Dtvcc)
 pub fn do_cb_dtvcc(ctx: &mut lib_cc_decode, dtvcc: &mut Dtvcc, cc_block: &[u8]) -> bool {
     let cc_valid = (cc_block[0] & 4) >> 2;
     let cc_type = cc_block[0] & 3;
@@ -295,8 +509,8 @@ pub fn do_cb_dtvcc(ctx: &mut lib_cc_decode, dtvcc: &mut Dtvcc, cc_block: &[u8])
     if ctx.write_format != ccx_output_format::CCX_OF_DVDRAW
         && ctx.write_format != ccx_output_format::CCX_OF_RAW
         && (cc_block[0] == 0xFA || cc_block[0] == 0xFC || cc_block[0] == 0xFD)
-        && (cc_block[1] & 0x7F) == 0
-        && (cc_block[2] & 0x7F) == 0
+        && (cc_block[1] & PARITY_BIT_MASK) == 0
+        && (cc_block[2] & PARITY_BIT_MASK) == 0
     {
         return true;
     }
@@ -337,6 +551,59 @@ pub fn do_cb_dtvcc(ctx: &mut lib_cc_decode, dtvcc: &mut Dtvcc, cc_block: &[u8])
     true
 }
 
+/// Process CC data according to its type (using DtvccRust - persistent context)
+pub fn do_cb_dtvcc_rust(ctx: &mut lib_cc_decode, dtvcc: &mut DtvccRust, cc_block: &[u8]) -> bool {
+    let cc_valid = (cc_block[0] & 4) >> 2;
+    let cc_type = cc_block[0] & 3;
+    let mut timeok = true;
+
+    if ctx.write_format != ccx_output_format::CCX_OF_DVDRAW
+        && ctx.write_format != ccx_output_format::CCX_OF_RAW
+        && (cc_block[0] == 0xFA || cc_block[0] == 0xFC || cc_block[0] == 0xFD)
+        && (cc_block[1] & PARITY_BIT_MASK) == 0
+        && (cc_block[2] & PARITY_BIT_MASK) == 0
+    {
+        return true;
+    }
+
+    if cc_valid == 1 || cc_type == 3 {
+        ctx.cc_stats[cc_type as usize] += 1;
+        match cc_type {
+            // Type 0 and 1 are for CEA-608 data. Handled by C code, do nothing
+            0 | 1 => {}
+            // Type 2 and 3 are for CEA-708 data.
+            2 | 3 => {
+                let current_time = if ctx.timing.is_null() {
+                    0
+                } else {
+                    unsafe { (*ctx.timing).get_fts(ctx.current_field as u8) }
+                };
+                ctx.current_field = 3;
+
+                // Check whether current time is within start and end bounds
+                if is_true(ctx.extraction_start.set)
+                    && current_time < ctx.extraction_start.time_in_ms
+                {
+                    timeok = false;
+                }
+                if is_true(ctx.extraction_end.set) && current_time > ctx.extraction_end.time_in_ms {
+                    timeok = false;
+                    ctx.processed_enough = 1;
+                }
+
+                if timeok && ctx.write_format != ccx_output_format::CCX_OF_RAW {
+                    dtvcc.process_cc_data(cc_valid, cc_type, cc_block[1], cc_block[2]);
+                }
+                // Note: cb_708 is incremented by the C code in do_cb(), not here.
+                // Previously incrementing here caused a double-increment bug that
+                // resulted in incorrect start timestamps.
+            }
+            _ => warn!("Invalid cc_type"),
+        }
+    }
+    true
+}
+
 /// Close a Windows handle by wrapping it in a File and dropping it.
 ///
 /// # Safety
@@ -370,12 +637,20 @@ extern "C" fn ccxr_close_handle(handle: RawHandle) {
 /// - Double-dash options (e.g., `--quiet`) are left unchanged
 /// - Single-letter short options (e.g., `-o`) are left unchanged
 /// - Non-option arguments (e.g., `file.ts`) are left unchanged
-/// - Numeric options (e.g., `-1`, `-12`) are left unchanged (these are valid short options)
+/// - Numeric options `-1`, `-2`, `-12` are converted to `--output-field=N` for CEA-608 field selection
 fn normalize_legacy_option(arg: String) -> String {
+    // Handle legacy numeric options for CEA-608 field extraction
+    // These map to --output-field which is the modern equivalent
+    match arg.as_str() {
+        "-1" => return "--output-field=1".to_string(),
+        "-2" => return "--output-field=2".to_string(),
+        "-12" => return "--output-field=12".to_string(),
+        _ => {}
+    }
+
     // Check if it's a single-dash option with multiple characters (e.g., -quiet)
     // but not a short option with a value (e.g., -o filename)
     // Single-letter options like -o, -s should be left unchanged
-    // Numeric options like -1, -12 should also be left unchanged
     if arg.starts_with('-')
         && !arg.starts_with("--")
         && arg.len() > 2
@@ -527,13 +802,22 @@ mod test {
         let mut cc_block = [0x15, 0x2F, 0x7D];
         assert!(validate_cc_pair(&mut cc_block));
         // Check for replaced bit when 1st byte doesn't pass parity
-        assert_eq!(cc_block[1], 0x7F);
+        assert_eq!(cc_block[1], CC_SOLID_BLANK);
 
         // Invalid CEA-608 data
         let mut cc_block = [0x15, 0x2F, 0x5E];
         assert!(!validate_cc_pair(&mut cc_block));
     }
 
+    #[test]
+    fn test_validate_cc_pair_invalid_length() {
+        let mut short = [0x97, 0x1F];
+        assert!(!validate_cc_pair(&mut short));
+
+        let mut long = [0x97, 0x1F, 0x3C, 0x00];
+        assert!(!validate_cc_pair(&mut long));
+    }
+
     #[test]
     fn test_do_cb() {
         let mut dtvcc_ctx = crate::decoder::test::initialize_dtvcc_ctx();
@@ -595,12 +879,18 @@ mod test {
 
     #[test]
     fn test_normalize_legacy_option_numeric_options() {
-        // Numeric options should remain unchanged (these are valid ccextractor options)
-        assert_eq!(normalize_legacy_option("-1".to_string()), "-1".to_string());
-        assert_eq!(normalize_legacy_option("-2".to_string()), "-2".to_string());
+        // Legacy numeric options for CEA-608 field selection are converted to --output-field
+        assert_eq!(
+            normalize_legacy_option("-1".to_string()),
+            "--output-field=1".to_string()
+        );
+        assert_eq!(
+            normalize_legacy_option("-2".to_string()),
+            "--output-field=2".to_string()
+        );
         assert_eq!(
             normalize_legacy_option("-12".to_string()),
-            "-12".to_string()
+            "--output-field=12".to_string()
         );
     }
 

src/rust/src/libccxr_exports/demuxer.rs

@@ -7,14 +7,22 @@ use crate::demuxer::common_types::{
 };
 use lib_ccxr::common::{Codec, Options, StreamMode, StreamType};
 use lib_ccxr::time::Timestamp;
-use std::alloc::{alloc_zeroed, Layout};
 use std::ffi::CStr;
 use std::os::raw::{c_char, c_int, c_uchar, c_uint, c_void};
 
-// External C function declarations
+/// Poison pattern used to detect uninitialized pointers (0xCD repeated).
+/// This pattern is commonly used by debug memory allocators.
+#[cfg(target_pointer_width = "64")]
+const POISON_PTR_PATTERN: usize = 0xcdcdcdcdcdcdcdcd;
+#[cfg(target_pointer_width = "32")]
+const POISON_PTR_PATTERN: usize = 0xcdcdcdcd;
+
 extern "C" {
     fn activity_input_file_closed();
     fn close(fd: c_int) -> c_int;
+    fn malloc(size: usize) -> *mut c_void;
+    fn free(ptr: *mut c_void);
+    fn calloc(nmemb: usize, size: usize) -> *mut c_void;
 }
 
 pub fn copy_c_array_to_rust_vec(
@@ -91,61 +99,89 @@ pub unsafe fn copy_demuxer_from_rust_to_c(c_demuxer: *mut ccx_demuxer, rust_demu
     c.global_timestamp_inited = rust_demuxer.global_timestamp_inited.millis() as c_int;
 
     // PID buffers - extra defensive version
+    // We iterate through all possible PIDs (up to 8191 for PSI) to ensure state synchronization.
+    // CRITICAL: We must free existing pointers in the C structure before overwriting them
+    // to prevent massive memory leaks during the demuxing process, as this function
+    // is called repeatedly to sync state between Rust and C.
     let pid_buffers_len = rust_demuxer.pid_buffers.len().min(8191);
-    for i in 0..pid_buffers_len {
-        let pid_buffer = rust_demuxer.pid_buffers[i];
-        if !pid_buffer.is_null() {
-            // Try to safely access the pointer
-            match std::panic::catch_unwind(|| unsafe { &*pid_buffer }) {
-                Ok(rust_psi) => {
-                    let c_psi = unsafe { rust_psi.to_ctype() };
-                    let c_ptr = Box::into_raw(Box::new(c_psi));
-                    c.PID_buffers[i] = c_ptr;
-                }
-                Err(_) => {
-                    // Pointer was invalid, set to null
-                    eprintln!("Warning: Invalid PID buffer pointer at index {i}");
-                    c.PID_buffers[i] = std::ptr::null_mut();
+    for i in 0..8191 {
+        // Free existing pointer if any.
+        // SAFETY: We use C's free to be compatible with memory that might be allocated by C.
+        // We also check for POISON_PTR_PATTERN for safety in debug builds.
+        if !c.PID_buffers[i].is_null() && c.PID_buffers[i] as usize != POISON_PTR_PATTERN {
+            unsafe {
+                free(c.PID_buffers[i] as *mut c_void);
+                c.PID_buffers[i] = std::ptr::null_mut();
+            }
+        }
+
+        if i < pid_buffers_len {
+            let pid_buffer = rust_demuxer.pid_buffers[i];
+            if !pid_buffer.is_null() {
+                // Try to safely access the pointer using catch_unwind to prevent
+                // a panic in Rust from crashing the entire C application.
+                // This is a defensive measure for FFI robustness.
+                match std::panic::catch_unwind(|| unsafe { &*pid_buffer }) {
+                    Ok(rust_psi) => {
+                        let c_psi = unsafe { rust_psi.to_ctype() };
+                        let c_ptr =
+                            unsafe { malloc(std::mem::size_of::<crate::bindings::PSI_buffer>()) }
+                                as *mut crate::bindings::PSI_buffer;
+                        if !c_ptr.is_null() {
+                            unsafe {
+                                std::ptr::write(c_ptr, c_psi);
+                            }
+                            c.PID_buffers[i] = c_ptr;
+                        }
+                    }
+                    Err(_) => {
+                        // Pointer was invalid, log and skip
+                        eprintln!("Warning: Invalid PID buffer pointer at index {i}");
+                    }
                 }
             }
-        } else {
-            c.PID_buffers[i] = std::ptr::null_mut();
         }
     }
 
-    // Clear remaining slots if rust array is smaller than C array
-    for i in pid_buffers_len..8191 {
-        c.PID_buffers[i] = std::ptr::null_mut();
-    }
-
     // PIDs programs - extra defensive version
+    // Similar to PID_buffers, we manage ownership of PMT entries.
+    // We check for POISON_PTR_PATTERN to avoid freeing uninitialized memory in debug builds.
     let pids_programs_len = rust_demuxer.pids_programs.len().min(65536);
-    for i in 0..pids_programs_len {
-        let pmt_entry = rust_demuxer.pids_programs[i];
-        if !pmt_entry.is_null() {
-            // Try to safely access the pointer
-            match std::panic::catch_unwind(|| unsafe { &*pmt_entry }) {
-                Ok(rust_pmt) => {
-                    let c_pmt = unsafe { rust_pmt.to_ctype() };
-                    let c_ptr = Box::into_raw(Box::new(c_pmt));
-                    c.PIDs_programs[i] = c_ptr;
-                }
-                Err(_) => {
-                    // Pointer was invalid, set to null
-                    eprintln!("Warning: Invalid PMT entry pointer at index {i}");
-                    c.PIDs_programs[i] = std::ptr::null_mut();
+    for i in 0..65536 {
+        // Free existing pointer if any and it's not a poison pattern.
+        // SAFETY: We use C's free to be compatible with memory that might be allocated by C.
+        if !c.PIDs_programs[i].is_null() && c.PIDs_programs[i] as usize != POISON_PTR_PATTERN {
+            unsafe {
+                free(c.PIDs_programs[i] as *mut c_void);
+                c.PIDs_programs[i] = std::ptr::null_mut();
+            }
+        }
+
+        if i < pids_programs_len {
+            let pmt_entry = rust_demuxer.pids_programs[i];
+            if !pmt_entry.is_null() {
+                // Safely convert and move ownership to C
+                match std::panic::catch_unwind(|| unsafe { &*pmt_entry }) {
+                    Ok(rust_pmt) => {
+                        let c_pmt = unsafe { rust_pmt.to_ctype() };
+                        let c_ptr =
+                            unsafe { malloc(std::mem::size_of::<crate::bindings::PMT_entry>()) }
+                                as *mut crate::bindings::PMT_entry;
+                        if !c_ptr.is_null() {
+                            unsafe {
+                                std::ptr::write(c_ptr, c_pmt);
+                            }
+                            c.PIDs_programs[i] = c_ptr;
+                        }
+                    }
+                    Err(_) => {
+                        eprintln!("Warning: Invalid PMT entry pointer at index {i}");
+                    }
                 }
             }
-        } else {
-            c.PIDs_programs[i] = std::ptr::null_mut();
         }
     }
 
-    // Clear remaining slots if rust array is smaller than C array
-    for i in pids_programs_len..65536 {
-        c.PIDs_programs[i] = std::ptr::null_mut();
-    }
-
     // PIDs seen array
     for (i, &val) in rust_demuxer.pids_seen.iter().take(65536).enumerate() {
         c.PIDs_seen[i] = val as c_int;
@@ -258,7 +294,15 @@ pub unsafe fn copy_demuxer_from_c_to_rust(ccx: *const ccx_demuxer) -> CcxDemuxer
             if buffer_ptr.is_null() {
                 None
             } else {
-                Some(Box::into_raw(Box::new(PSIBuffer::from_ctype(*buffer_ptr)?)))
+                let rust_item = PSIBuffer::from_ctype(*buffer_ptr)?;
+                let rust_ptr =
+                    unsafe { malloc(std::mem::size_of::<PSIBuffer>()) } as *mut PSIBuffer;
+                if !rust_ptr.is_null() {
+                    unsafe {
+                        std::ptr::write(rust_ptr, rust_item);
+                    }
+                }
+                Some(rust_ptr)
             }
         })
         .collect::<Vec<_>>();
@@ -266,10 +310,17 @@ pub unsafe fn copy_demuxer_from_c_to_rust(ccx: *const ccx_demuxer) -> CcxDemuxer
         .PIDs_programs
         .iter()
         .filter_map(|&buffer_ptr| {
-            if buffer_ptr.is_null() || buffer_ptr as usize == 0xcdcdcdcdcdcdcdcd {
+            if buffer_ptr.is_null() || buffer_ptr as usize == POISON_PTR_PATTERN {
                 None
             } else {
-                Some(Box::into_raw(Box::new(PMTEntry::from_ctype(*buffer_ptr)?)))
+                let rust_item = PMTEntry::from_ctype(*buffer_ptr)?;
+                let rust_ptr = unsafe { malloc(std::mem::size_of::<PMTEntry>()) } as *mut PMTEntry;
+                if !rust_ptr.is_null() {
+                    unsafe {
+                        std::ptr::write(rust_ptr, rust_item);
+                    }
+                }
+                Some(rust_ptr)
             }
         })
         .collect::<Vec<_>>();
@@ -360,8 +411,7 @@ pub unsafe fn copy_demuxer_from_c_to_rust(ccx: *const ccx_demuxer) -> CcxDemuxer
 ///
 /// This function is unsafe because we are calling a C struct and using alloc_zeroed to initialize it.
 pub unsafe fn alloc_new_demuxer() -> *mut ccx_demuxer {
-    let layout = Layout::new::<ccx_demuxer>();
-    let ptr = alloc_zeroed(layout) as *mut ccx_demuxer;
+    let ptr = calloc(1, std::mem::size_of::<ccx_demuxer>()) as *mut ccx_demuxer;
 
     if ptr.is_null() {
         panic!("Failed to allocate memory for ccx_demuxer");

src/rust/src/libccxr_exports/mod.rs

@@ -33,10 +33,11 @@ pub unsafe extern "C" fn ccxr_init_basic_logger() {
         .unwrap_or(DebugMessageFlag::VERBOSE);
     let mask = DebugMessageMask::new(debug_mask, debug_mask_on_debug);
     let gui_mode_reports = ccx_options.gui_mode_reports != 0;
+    // CCX_MESSAGES_QUIET=0, CCX_MESSAGES_STDOUT=1, CCX_MESSAGES_STDERR=2
     let messages_target = match ccx_options.messages_target {
-        0 => OutputTarget::Stdout,
-        1 => OutputTarget::Stderr,
-        2 => OutputTarget::Quiet,
+        0 => OutputTarget::Quiet,
+        1 => OutputTarget::Stdout,
+        2 => OutputTarget::Stderr,
         _ => OutputTarget::Stderr, // Default to stderr for invalid values
     };
     let _ = set_logger(CCExtractorLogger::new(
@@ -46,6 +47,28 @@ pub unsafe extern "C" fn ccxr_init_basic_logger() {
     ));
 }
 
+/// Updates the logger target after command-line arguments have been parsed.
+/// This is needed because the logger is initialized before argument parsing,
+/// and options like --quiet need to be applied afterwards.
+///
+/// # Safety
+///
+/// `ccx_options` in C must be properly initialized and the logger must have
+/// been initialized via `ccxr_init_basic_logger` before calling this function.
+#[no_mangle]
+pub unsafe extern "C" fn ccxr_update_logger_target() {
+    // CCX_MESSAGES_QUIET=0, CCX_MESSAGES_STDOUT=1, CCX_MESSAGES_STDERR=2
+    let messages_target = match ccx_options.messages_target {
+        0 => OutputTarget::Quiet,
+        1 => OutputTarget::Stdout,
+        2 => OutputTarget::Stderr,
+        _ => OutputTarget::Stderr,
+    };
+    if let Some(mut logger) = logger_mut() {
+        logger.set_target(messages_target);
+    }
+}
+
 /// Rust equivalent for `verify_crc32` function in C. Uses C-native types as input and output.
 ///
 /// # Safety
@@ -54,6 +77,10 @@ pub unsafe extern "C" fn ccxr_init_basic_logger() {
 /// or less than `len`.
 #[no_mangle]
 pub unsafe extern "C" fn ccxr_verify_crc32(buf: *const u8, len: c_int) -> c_int {
+    // Safety: avoid NULL pointer and negative length causing usize wraparound
+    if buf.is_null() || len < 0 {
+        return 0;
+    }
     let buf = std::slice::from_raw_parts(buf, len as usize);
     if verify_crc32(buf) {
         1

windows/ccextractor.sln

@@ -19,4 +19,4 @@ Global
 	GlobalSection(SolutionProperties) = preSolution
 		HideSolutionNode = FALSE
 	EndGlobalSection
-EndGlobal
+EndGlobal