starred/terminal
1
0
Fork 0
mirror of https://github.com/microsoft/terminal.git synced 2026-02-04 05:35:20 +00:00
Code Issues 19.0k Packages Projects Releases 20 Wiki Activity

cygwin with win32-raw-mode enabled, console consumes all available memory. #22184

New Issue
Closed
opened 2026-01-31 08:05:53 +00:00 by claunia · 7 comments
claunia commented 2026-01-31 08:05:53 +00:00
Owner
Copy Link

Originally created by @adamyg on GitHub (Aug 29, 2024).

Windows Terminal version

1.20.11781.0

Windows build number

10.0.19045.4780

Other Software

cygwin64, either

  • console shell running under "c:\cygwin64\bin\bash --login"; or
  • run rawtest application with cygwin.dll in path.

Steps to reproduce

Application
enables win32-input-mode (\033[?9001h), fast key input triggers condition; see attached.
memory-usage

Test application:

/*
 *  win32-input-mode key test
 *
 *  Build: gcc -o conkeytest conkeytest.c
 */

#include <termios.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>

static void hex(const void *buf, unsigned len);

int
main(void)
{
        struct termios bak, cfg;
        char buf[128];
        int cnt, q;

        tcgetattr(STDIN_FILENO, &cfg);
        bak = cfg;
        cfmakeraw(&cfg);
        cfg.c_lflag &= ~ECHO;
        tcsetattr(STDIN_FILENO, TCSADRAIN, &cfg);

        printf("\033[?9001h");      // enable win32-input-mode.
//      printf("\033[?1002h");      // enable cell-motion tracking tracking.
//      printf("\033[?1006h");      // enable SGR extended mouse mode.
        printf("\n\r");
        fflush(stdout);

        for (q = 0; (cnt = read(STDIN_FILENO, buf, sizeof(buf) - 1)) > 0;) {
                buf[cnt] = 0;
                if (0 == strcmp(buf, "\x1b[81;16;113;1;0;1_")) { // q-down
                        if (++q == 2)
                                break;
                } else if (0 == strcmp(buf, "\x1b[81;16;113;0;0;1_")) { // q-up
                } else {
                        q = 0;
                }
                hex(buf, cnt);
        }

//      printf("\033[?1006l");      // disable SGR extended mouse mode.
//      printf("\033[?1002l");      // disable cell-motion tracking tracking.
        printf("\033[?9001l");      // disable win32-input-mode.
        fflush(stdout);

        tcsetattr(STDIN_FILENO, TCSADRAIN, &bak);
}


static void
hex(const void *buf, unsigned len)
{
#define HEXWIDTH 32
#define HEXCOLUMN 4
        const unsigned char *cursor = buf, *end = cursor + len;
        unsigned offset = 0;

        for (offset = 0; cursor < end; offset += HEXWIDTH) {
                const unsigned char *data = cursor;

                printf("%04x: ", offset);
                for (unsigned col = 0; col < HEXWIDTH; ++col) {
                        if ((col % HEXCOLUMN) == 0 && col)
                                printf(" |");
                        if (data == end) {
                                printf("   ");
                        } else {
                                printf(" %02x", *data++);
                        }
                }
                printf("   ");
                for (unsigned col = 0; col < HEXWIDTH; ++col) {
                        if (cursor == end) {
                                printf("%*s", HEXWIDTH - col, "");
                                break;
                        }
                        const unsigned char c = *cursor++;
                        printf("%c", (c >= ' ' && c < 0x7f ? c : '.'));
                }
                printf("\n\r");
        }
        fflush(stdout);
}

//end

Expected Behavior

Raw keys code reported without issue.
example-run

Actual Behavior

OpenConsole rapidly consumes memory and cpu, shall consume all available resources unless killed;
application becomes unresponsive.

memory-usage

Originally created by @adamyg on GitHub (Aug 29, 2024). ### Windows Terminal version 1.20.11781.0 ### Windows build number 10.0.19045.4780 ### Other Software cygwin64, either - console shell running under "c:\cygwin64\bin\bash --login"; or - run rawtest application with cygwin.dll in path. ### Steps to reproduce Application enables win32-input-mode (\033[?9001h), fast key input triggers condition; see attached. ![memory-usage](https://github.com/user-attachments/assets/4d0e94ff-000e-4ba4-b8a4-2e4f07f7cf1d) Test application: ``` /* * win32-input-mode key test * * Build: gcc -o conkeytest conkeytest.c */ #include <termios.h> #include <stdio.h> #include <stdlib.h> #include <string.h> #include <unistd.h> static void hex(const void *buf, unsigned len); int main(void) { struct termios bak, cfg; char buf[128]; int cnt, q; tcgetattr(STDIN_FILENO, &cfg); bak = cfg; cfmakeraw(&cfg); cfg.c_lflag &= ~ECHO; tcsetattr(STDIN_FILENO, TCSADRAIN, &cfg); printf("\033[?9001h"); // enable win32-input-mode. // printf("\033[?1002h"); // enable cell-motion tracking tracking. // printf("\033[?1006h"); // enable SGR extended mouse mode. printf("\n\r"); fflush(stdout); for (q = 0; (cnt = read(STDIN_FILENO, buf, sizeof(buf) - 1)) > 0;) { buf[cnt] = 0; if (0 == strcmp(buf, "\x1b[81;16;113;1;0;1_")) { // q-down if (++q == 2) break; } else if (0 == strcmp(buf, "\x1b[81;16;113;0;0;1_")) { // q-up } else { q = 0; } hex(buf, cnt); } // printf("\033[?1006l"); // disable SGR extended mouse mode. // printf("\033[?1002l"); // disable cell-motion tracking tracking. printf("\033[?9001l"); // disable win32-input-mode. fflush(stdout); tcsetattr(STDIN_FILENO, TCSADRAIN, &bak); } static void hex(const void *buf, unsigned len) { #define HEXWIDTH 32 #define HEXCOLUMN 4 const unsigned char *cursor = buf, *end = cursor + len; unsigned offset = 0; for (offset = 0; cursor < end; offset += HEXWIDTH) { const unsigned char *data = cursor; printf("%04x: ", offset); for (unsigned col = 0; col < HEXWIDTH; ++col) { if ((col % HEXCOLUMN) == 0 && col) printf(" |"); if (data == end) { printf(" "); } else { printf(" %02x", *data++); } } printf(" "); for (unsigned col = 0; col < HEXWIDTH; ++col) { if (cursor == end) { printf("%*s", HEXWIDTH - col, ""); break; } const unsigned char c = *cursor++; printf("%c", (c >= ' ' && c < 0x7f ? c : '.')); } printf("\n\r"); } fflush(stdout); } //end ``` ### Expected Behavior Raw keys code reported without issue. ![example-run](https://github.com/user-attachments/assets/dc1b98fb-d239-43d3-8076-7b3d1c57f546) ### Actual Behavior OpenConsole rapidly consumes memory and cpu, shall consume all available resources unless killed; application becomes unresponsive. ![memory-usage](https://github.com/user-attachments/assets/f40dfa39-bec1-4865-a5ad-514a0f92424e)
claunia added the Needs-TriageIssue-BugResolution-External labels 2026-01-31 08:05:53 +00:00
claunia commented 2026-01-31 08:05:54 +00:00
Author
Owner
Copy Link

@lhecker commented on GitHub (Aug 29, 2024):

Using your test application, I was easily able to reproduce the issue by pasting a few lines of text (>100 characters) 5-10 times. It would then go out of control as you described.

I believe this is a bug in cygwin. When the issue occurs, something calls WriteConsoleInput. In my case with 1340 INPUT_RECORDs which we would then store and offer for reading. Then WriteConsoleInput would get called with 20100 records, 295356 records, and so on. It's definitely not a deadlock that's internal to OpenConsole/conhost. You may have to open a bug report for cygwin. At least to me it doesn't seem like there's a bug in your example code. We can't fix it either, because if someone calls WriteConsoleInput, we'll respect that request.

(The reason it may seem like a deadlock in OpenConsole is because our input processing is just extremely slow. The input buffer implementation was unfortunately "improved" by someone many years ago in the pursuit of OOP cleanliness. We're planning to fix that soon. But making it faster will not fix this bug, it'll simply make it run OOM faster. 😄)

@lhecker commented on GitHub (Aug 29, 2024): Using your test application, I was easily able to reproduce the issue by pasting a few lines of text (>100 characters) 5-10 times. It would then go out of control as you described. I believe this is a bug in cygwin. When the issue occurs, something calls `WriteConsoleInput`. In my case with 1340 `INPUT_RECORD`s which we would then store and offer for reading. Then `WriteConsoleInput` would get called with 20100 records, 295356 records, and so on. It's definitely not a deadlock that's internal to OpenConsole/conhost. You may have to open a bug report for cygwin. At least to me it doesn't seem like there's a bug in your example code. We can't fix it either, because if someone calls `WriteConsoleInput`, we'll respect that request. (The reason it may seem like a deadlock in OpenConsole is because our input processing is just extremely slow. The input buffer implementation was unfortunately "improved" by someone many years ago in the pursuit of OOP cleanliness. We're planning to fix that soon. But making it faster will not fix this bug, it'll simply make it run OOM faster. 😄)
claunia commented 2026-01-31 08:05:54 +00:00
Author
Owner
Copy Link

@adamyg commented on GitHub (Aug 29, 2024):

Shall continue researching, might review Cygwin's console interface for any insight.
One suspect is the fhandler_console::cons_master_thread (), which attempts to mine signals within the input strem; yet it may not handle a single key being represented by multiple events.; furthermore one of the few uses of WriteConsoleInput

Note: upon disabling win32-input-mode, the problem does not not occur.

Assume there are no plans on supporting Cygwin's equiv cygwin-raw-mode, ESC[?2000h, generates ESC{Kd;Rc;Vk;Sc;Uc;CsK

@adamyg commented on GitHub (Aug 29, 2024): Shall continue researching, might review Cygwin's console interface for any insight. _One suspect is the fhandler_console::cons_master_thread (), which attempts to mine signals within the input strem; yet it may not handle a single key being represented by multiple events._; furthermore one of the few uses of WriteConsoleInput Note: upon disabling win32-input-mode, the problem does not not occur. Assume there are no plans on supporting Cygwin's equiv cygwin-raw-mode, ``ESC[?2000h``, generates ``ESC{Kd;Rc;Vk;Sc;Uc;CsK``
claunia commented 2026-01-31 08:05:55 +00:00
Author
Owner
Copy Link

@lhecker commented on GitHub (Aug 29, 2024):

Currently we have no plans adding support for other input protocols ourselves, but we'd gladly accept any PRs that add support for them!

@lhecker commented on GitHub (Aug 29, 2024): Currently we have no plans adding support for other input protocols ourselves, but we'd gladly accept any PRs that add support for them!
claunia commented 2026-01-31 08:05:55 +00:00
Author
Owner
Copy Link

@adamyg commented on GitHub (Aug 31, 2024):

Discussion raised within cygwin mail group.
One suspect is the fhandler_console::cons_master_thread (), which attempts to mine signals within the input stream;
yet it may not handle a single key being represented by multiple events.; furthermore one of the few uses of WriteConsoleInput. cons_master_thread peeks at the input event, removing signal events and pushes back others.

Please confirm the intended behavior of WriteConsoleInput() when win32-input-mode, is enabled?

Wondering if its behavior should be to not expand key events without Vk=0/Sc=0 values, from research, these represent the expanded pseudo key events. This behavior may also address the double encoding of mouse event under win32-input-mode,

@adamyg commented on GitHub (Aug 31, 2024): Discussion raised within cygwin mail group. _One suspect is the fhandler_console::cons_master_thread (), which attempts to mine signals within the input stream; yet it may not handle a single key being represented by multiple events.; furthermore one of the few uses of WriteConsoleInput. cons_master_thread peeks at the input event, removing signal events and pushes back others._ Please confirm the intended behavior of _WriteConsoleInput()_ when _win32-input-mode_, is enabled? Wondering if its behavior should be to not expand key events without Vk=0/Sc=0 values, from research, these represent the expanded pseudo key events. This behavior may also address the double encoding of mouse event under _win32-input-mode_,
claunia commented 2026-01-31 08:05:55 +00:00
Author
Owner
Copy Link

@adamyg commented on GitHub (Aug 31, 2024):

Update: https://sourceware.org/pipermail/cygwin-patches/2024q3/012748.html

@adamyg commented on GitHub (Aug 31, 2024): Update: https://sourceware.org/pipermail/cygwin-patches/2024q3/012748.html
claunia commented 2026-01-31 08:05:55 +00:00
Author
Owner
Copy Link

@lhecker commented on GitHub (Sep 3, 2024):

I'm glad that you were able to fix the issue in cygwin. I'll be closing the issue then. Thank you for reporting it in their mailing list as well as fixing it! 😻

It may be worth nothing that the fix will be required even long-term. If an application requests win32-input-mode (w32im), then we'll have to respect this wish and translate any inputs to the w32im sequences, even if they come from WriteConsoleInput. This is the only way we can losslessly transmit INPUT_RECORDs across SSH for instance. As such this is an issue we cannot fix on our side unfortunately.

@lhecker commented on GitHub (Sep 3, 2024): I'm glad that you were able to fix the issue in cygwin. I'll be closing the issue then. Thank you for reporting it in their mailing list as well as fixing it! 😻 It may be worth nothing that the fix will be required even long-term. If an application requests win32-input-mode (w32im), then we'll have to respect this wish and translate _any_ inputs to the w32im sequences, even if they come from `WriteConsoleInput`. This is the only way we can losslessly transmit `INPUT_RECORD`s across SSH for instance. As such this is an issue we cannot fix on our side unfortunately.
claunia commented 2026-01-31 08:05:55 +00:00
Author
Owner
Copy Link

@lhecker commented on GitHub (Sep 3, 2024):

Oh, I missed your question:

Wondering if its behavior should be to not expand key events without Vk=0/Sc=0 values, from research, these represent the expanded pseudo key events. This behavior may also address the double encoding of mouse event under win32-input-mode,

The intention of the protocol is to be absolutely lossless when it comes to INPUT_RECORD. So even if wVirtualKeyCode and wVirtualScanCode are zero, members like dwControlKeyState may not be zero and we want to preserve everything perfectly.

Edit: I can confirm that mouse events are being double-encoded. I've opened #17851 to track the issue. I'll fix that ASAP and see if we can backport it.

@lhecker commented on GitHub (Sep 3, 2024): Oh, I missed your question: > Wondering if its behavior should be to not expand key events without Vk=0/Sc=0 values, from research, these represent the expanded pseudo key events. This behavior may also address the double encoding of mouse event under _win32-input-mode_, The intention of the protocol is to be absolutely lossless when it comes to `INPUT_RECORD`. So even if `wVirtualKeyCode` and `wVirtualScanCode` are zero, members like `dwControlKeyState` may not be zero and we want to preserve everything perfectly. Edit: I can confirm that mouse events are being double-encoded. I've opened #17851 to track the issue. I'll fix that ASAP and see if we can backport it.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
⛺ Reserved
A11yCO
A11yMAS
A11ySev1
A11ySev2
A11ySev3
A11yTTValidated
A11yUsable
A11yVoiceAccess
A11yWCAG
Area-Accessibility
Area-AtlasEngine
Area-AzureShell
Area-Build
Area-Build
Area-Chat
Area-CmdPal
Area-CodeHealth
Area-Commandline
Area-CookedRead
Area-DefApp
Area-Extensibility
Area-Fonts
Area-GroupPolicy
Area-i18n
Area-Input
Area-Interaction
Area-Interop
Area-Localization
Area-Output
Area-Performance
Area-Portable
Area-Quality
Area-Remoting
Area-Rendering
Area-Schema
Area-Server
Area-Settings
Area-SettingsUI
Area-ShellExtension
Area-ShellExtension
Area-ShellExtension
Area-Suggestions
Area-Suggestions
Area-TerminalConnection
Area-TerminalControl
Area-Theming
Area-UserInterface
Area-VT
Area-Windowing
Area-WPFControl
AutoMerge
Blocking-Ingestion
Culprit-Centennial
Culprit-WinUI
Disability-All
Disability-Blind
Disability-LowVision
Disability-Mobility
External-Blocked-WinUI3
Fixed
Gathering-Data
good first issue
HCL-E+D
HCL-WindowsTerminal
Help Wanted
Impact-Compatibility
Impact-Compliance
Impact-Correctness
Impact-Visual
In-PR
InclusionBacklog
InclusionBacklog-Windows TerminalWin32
InclusionCommitted-202206
Issue-Bug
Issue-Docs
Issue-Feature
Issue-Feature
Issue-Question
Issue-Samples
Issue-Scenario
Issue-Task
Needs-Attention
Needs-Author-Feedback
Needs-Bisect
Needs-Discussion
Needs-Repro
Needs-Tag-Fix
Needs-Tag-Fix
Needs-Triage
No-Recent-Activity
Priority-0
Priority-1
Priority-2
Priority-3
Product-Cmd.exe
Product-Colortool
Product-Colortool
Product-Colortool
Product-Conhost
Product-Conpty
Product-Meta
Product-Powershell
Product-Terminal
Product-WSL
pull-request
Mirrored from GitHub Pull Request
 Resolution-Answered
Resolution-By-Design
Resolution-Duplicate
Resolution-External
Resolution-Fix-Available
Resolution-Fix-Committed
Resolution-No-Repro
Resolution-Won't-Fix
Severity-Blocking
Severity-Crash
Severity-DataLoss
spam
this-will-be-a-breaking-change
Tracking-External
WindowsTerminal_Win32
Work-Item
zAskModeBug
zInbox-Bug
No Label Issue-Bug Needs-Triage Resolution-External
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
claunia
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: starred/terminal#22184
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?