Echoing to stderr does not work in git hooks anymore

[t-b]

Existing issues matching what you're seeing

• I was not able to find an open or closed issue matching what I'm seeing

Git for Windows version

git version 2.53.0.rc0.windows.1
cpu: x86_64
built from commit: 402505fde7dac415cd5437aa0e57b9b4d9efee4c
sizeof-long: 4
sizeof-size_t: 8
shell-path: D:/git-sdk-64-build-installers/usr/bin/sh
rust: disabled
feature: fsmonitor--daemon
libcurl: 8.18.0
OpenSSL: OpenSSL 3.5.4 30 Sep 2025
zlib: 1.3.1
SHA-1: SHA1_DC
SHA-256: SHA256_BLK
default-ref-format: files
default-hash: sha1

Windows version

Windows 11

Windows CPU architecture

x86_64 (64-bit)

Additional Windows version information

Microsoft Windows [Version 10.0.26200.7623]
(c) Microsoft Corporation. Alle Rechte vorbehalten.

Options set during installation

Editor Option: VIM
Custom Editor Path:
Default Branch Option: main
Path Option: CmdTools
SSH Option: OpenSSH
Tortoise Option: false
CURL Option: OpenSSL
CRLF Option: CRLFCommitAsIs
Bash Terminal Option: ConHost
Git Pull Behavior Option: Rebase
Use Credential Manager: Disabled
Performance Tweaks FSCache: Disabled
Enable Symlinks: Disabled
Enable FSMonitor: Disabled

Other interesting things

No response

Terminal/shell

Bash in Windows Terminal

Commands that trigger the issue

$ cat .git/hooks/pre-push

#!/bin/bash

set -e

echo "abcd" > /dev/stderr

git push

Expected behaviour

This worked with 2.52.0. From 1 my uneducated guess would be that

When calling Microsoft Store apps, their standard I/O is now git-for-windows/msys2-runtime#122 (meaning: You can call an interactive Python interpreter without the winpty hack mentioned in the release notes' Known Issues).

is related.

Doing echo "abcd" > /dev/stderr in the terminal directly works.

Actual behaviour

.git/hooks/pre-push: line 5: /dev/stderr: No such file or directory
error: failed to push some refs to 'XXX'

Repository

No response

[rimrul]

There is also some recent discussion about the pre-push hook and stdout to stderr redirection on the mailinglist recently. That might also be related, but I haven't kept up with the details.

[t-b]

@rimrul Are you thinking of https://lore.kernel.org/git/ab578804-891e-edcc-12a6-8b1030d1bacb@apache.org?

[rimrul]

Yes. I think there was also a follow up patch series to it, but that probably hasn't made it into rc0.

[dscho]

@t-b could you please write a patch that adds a regression test? Or maybe find out whether the follow-up patch series has that already? Then I'll make sure that it gets into -rc1, at least for Git for Windows.

[t-b]

@t-b could you please write a patch that adds a regression test? Or maybe find out whether the follow-up patch series has that already? Then I'll make sure that it gets into -rc1, at least for Git for Windows.

I'll try to find some time for that.

[t-b]

Attached is a test 0001-t-t5571-prep-push-hook.sh-Add-test-with-writing-to-s.patch.

I've developed it on Linux and there v2.53-rc0 and latest master ab689ea (Revert "Merge branch 'cs/rebased-subtree-split'", 2026-01-25) have this test passing and also don't show the regression with my test case above.

[t-b]

The regression is gone in 2.53.0.rc1.windows.1 on Windows.

[t-b]

On Windows (using the very nicely done sdk!) I get a failure with 2.53.0.rc0.windows.1.1.g6c36579716 and a passing test with rc1.

[dscho]

On Windows (using the very nicely done sdk!) I get a failure with 2.53.0.rc0.windows.1.1.g6c36579716 and a passing test with rc1.

Awesome!

Do you want the regression test you developed to be part of -rc2 (and all subsequent Git for Windows versions)?

[t-b]

Do you want the regression test you developed to be part of -rc2 (and all subsequent Git for Windows versions)?

Sure. I did the obvious thing, see #6063.