Hot fixes from Git for Windows v2.49.0-rc0 #1867
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
/submit |
|
Submitted as [email protected] To fetch this version into To fetch this version to local tag |
![gitgitgadget[bot]](https://avatars.githubusercontent.com/in/12836?s=60&v=4){kind=small}
| } | ||
| return pw; | ||
| } | ||
|
|
There was a problem hiding this comment.
On the Git mailing list, Junio C Hamano wrote (reply to this):
"Johannes Schindelin via GitGitGadget" <[email protected]>
writes:
> From: Johannes Schindelin <[email protected]>
>
> In 590e081dea7c (ident: add NO_GECOS_IN_PWENT for systems without
> pw_gecos in struct passwd, 2011-05-19), code was introduced to iterate
> over the `gw_gecos` field; The loop variable is of type `char *`, which
> assumes that `gw_gecos` is writable.
>
> However, it is not necessarily writable (and it is a bad idea to have it
> writable in the first place), so let's switch the loop variable type to
> `const char *`.
>
> Signed-off-by: Johannes Schindelin <[email protected]>
> ---
> ident.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
Can we have a bit more details here (not in the commit but after the
three-dash line, meant for context) why the change from 2011 needs a
"hotfix" today? Has something in the build environment change? An
updated header file has changed definition for "struct passwd", or
something silly like that? Even though POSIX.1 does not define
gecos in its struct passwd in <pwd.h>, other string members like
pw_name and pw_dir are writable, which I found funny, and makes me
puzzled why this code from 2011 needs a hotfix all of the sudden.
> diff --git a/ident.c b/ident.c
> index caf41fb2a98..967895d8850 100644
> --- a/ident.c
> +++ b/ident.c
> @@ -59,7 +59,7 @@ static struct passwd *xgetpwuid_self(int *is_bogus)
>
> static void copy_gecos(const struct passwd *w, struct strbuf *name)
> {
> - char *src;
> + const char *src;
>
> /* Traditionally GECOS field had office phone numbers etc, separated
> * with commas. Also & stands for capitalized form of the login name.
The patch text itself looks perfectly fine, so I am not opposed to
its eventual application. Even though we do declare "src" as
non-const, we only use it to read from it, so declaring it as const
pointer is perfectly fine and more appropriate.
But I do not see any urgency relative to Git 2.48.0 (or Git 2.48.1
for that matter) to mark this as "hotfix" implying that it should be
included in 2.49-rc1 in either the patch or the proposed commit log
message.
Thanks.
| 'compat/win32/path-utils.c', | ||
| ] | ||
| elif host_machine.system() == 'windows' | ||
| libgit_sources += [ |
There was a problem hiding this comment.
On the Git mailing list, Patrick Steinhardt wrote (reply to this):
On Thu, Feb 27, 2025 at 03:44:09PM +0000, Johannes Schindelin via GitGitGadget wrote:
> From: Johannes Schindelin <[email protected]>
>
> In 904339edbd80 (Introduce support for the Meson build system,
> 2024-12-06) the `meson.build` file was introduced, adding also a
> Windows-specific list of source files. This list was obviously meant to
> be sorted alphabetically, but there is one mistake. Let's fix that.
This looks obviously good to me, thanks!
Patrick
|
User |
In 590e081 (ident: add NO_GECOS_IN_PWENT for systems without pw_gecos in struct passwd, 2011-05-19), code was introduced to iterate over the `gw_gecos` field; The loop variable is of type `char *`, which assumes that `gw_gecos` is writable. However, it is not necessarily writable (and it is a bad idea to have it writable in the first place), so let's switch the loop variable type to `const char *`. This is not a new problem, but what is new is the Meson build. While it does not trigger in CI builds, imitating the commands of `ci/run-build-and-tests.sh` in a regular Git for Windows SDK (`meson setup build . --fatal-meson-warnings --warnlevel 2 --werror --wrap-mode nofallback -Dfuzzers=true` followed by `meson compile -C build --` results in this beautiful error: "cc" [...] -o libgit.a.p/ident.c.obj "-c" ../ident.c ../ident.c: In function 'copy_gecos': ../ident.c:68:18: error: assignment discards 'const' qualifier from pointer target type [-Werror=discarded-qualifiers] 68 | for (src = get_gecos(w); *src && *src != ','; src++) { | ^ cc1.exe: all warnings being treated as errors Now, why does this not trigger in CI? The answer is as simple as it is puzzling: The `win+Meson` job completely side-steps Git for Windows' development environment, opting instead to use the GCC that is on the `PATH` in GitHub-hosted `windows-latest` runners. That GCC is pinned to v12.2.0 and targets the UCRT (unlikely to change any time soon, see https://github.com/actions/runner-images/blob/win25/20250303.1/images/windows/toolsets/toolset-2022.json#L132-L141). That is in stark contrast to Git for Windows, which uses GCC v14.2.0 and targets MSVCRT. Git for Windows' `Makefile`-based build also obviously uses different compiler flags, otherwise this compile error would have had plenty of opportunity in almost 14 years to surface. In other words, contrary to my expectations, the `win+Meson` job is ill-equipped to replace the `win build` job because it exercises a completely different tool version/compiler flags vector than what Git for Windows needs. Nevertheless, there is currently this huge push, including breaking changes after -rc1 and all, for switching to Meson. Therefore, we need to make it work, somehow, even in Git for Windows' SDK, hence this patch, at this point in time. Signed-off-by: Johannes Schindelin <[email protected]>
In 904339e (Introduce support for the Meson build system, 2024-12-06) the `meson.build` file was introduced, adding also a Windows-specific list of source files. This list was obviously meant to be sorted alphabetically, but there is one mistake. Let's fix that. Signed-off-by: Johannes Schindelin <[email protected]>
A late-comer to the v2.49.0 party, `sk/unit-test-oid`, added yet another array item to `CLAR_TEST_OBJS`, causing the `win+VS build` job to fail with symptoms like this one: unit-tests-lib.lib(u-oid-array.obj) : error LNK2019: unresolved external symbol cl_parse_any_oid referenced in function fill_array This is a similar scenario to the one that forced me to write 8afda42 (cmake: generalize the handling of the `UNIT_TEST_OBJS` list, 2024-09-18): The hard-coded echo of `CLAR_TEST_OBJS` in `CMakeLists.txt` that recapitulates faithfully what was already hard-coded in `Makefile` would either have to be updated whack-a-mole style, or generalized. Just like I chose the latter option for `UNIT_TEST_OBJS`, I now do the same for `CLAR_TEST_OBJS`. Signed-off-by: Johannes Schindelin <[email protected]>
|
/submit |
|
Submitted as [email protected] To fetch this version into To fetch this version to local tag |
| } | ||
| return pw; | ||
| } | ||
|
|
There was a problem hiding this comment.
On the Git mailing list, Patrick Steinhardt wrote (reply to this):
On Thu, Mar 06, 2025 at 10:26:18AM +0000, Johannes Schindelin via GitGitGadget wrote:
> From: Johannes Schindelin <[email protected]>
> Now, why does this not trigger in CI? The answer is as simple as it is
> puzzling: The `win+Meson` job completely side-steps Git for Windows'
> development environment, opting instead to use the GCC that is on the
> `PATH` in GitHub-hosted `windows-latest` runners. That GCC is pinned to
> v12.2.0 and targets the UCRT (unlikely to change any time soon, see
> https://github.com/actions/runner-images/blob/win25/20250303.1/images/windows/toolsets/toolset-2022.json#L132-L141).
> That is in stark contrast to Git for Windows, which uses GCC v14.2.0 and
> targets MSVCRT. Git for Windows' `Makefile`-based build also obviously
> uses different compiler flags, otherwise this compile error would have
> had plenty of opportunity in almost 14 years to surface.
Oh, interesting. I didn't even know that the Windows runners had GCC in
their PATH, and thus I didn't expect it to use that compiler at all. On
GitLab for example we can see that it uses the MSVC compiler as I did
expect [1]:
Activating VS 17.10.2
C compiler for the host machine: cl (msvc 19.40.33811 "Microsoft (R) C/C++ Optimizing Compiler Version 19.40.33811 for x64")
C linker for the host machine: link link 14.40.33811.0
But you're right, on GitHub that's not the case:
C compiler for the host machine: gcc (gcc 12.2.0 "gcc (x86_64-posix-seh-rev2, Built by MinGW-W64 project) 12.2.0")
C linker for the host machine: gcc ld.bfd 2.39
We can easily fix that by passing the `--vsenv` flag to Meson. I'll send
a patch soonish.
Patrick
[1]: https://gitlab.com/gitlab-org/git/-/jobs/9324989037#L95
[2]: https://github.com/git/git/actions/runs/13686408338/job/38270746786#step:5:15There was a problem hiding this comment.
On the Git mailing list, Johannes Schindelin wrote (reply to this):
Hi Patrick,
On Thu, 6 Mar 2025, Patrick Steinhardt wrote:
> On Thu, Mar 06, 2025 at 10:26:18AM +0000, Johannes Schindelin via GitGitGadget wrote:
> > From: Johannes Schindelin <[email protected]>
> > Now, why does this not trigger in CI? The answer is as simple as it is
> > puzzling: The `win+Meson` job completely side-steps Git for Windows'
> > development environment, opting instead to use the GCC that is on the
> > `PATH` in GitHub-hosted `windows-latest` runners. That GCC is pinned to
> > v12.2.0 and targets the UCRT (unlikely to change any time soon, see
> > https://github.com/actions/runner-images/blob/win25/20250303.1/images/windows/toolsets/toolset-2022.json#L132-L141).
> > That is in stark contrast to Git for Windows, which uses GCC v14.2.0 and
> > targets MSVCRT. Git for Windows' `Makefile`-based build also obviously
> > uses different compiler flags, otherwise this compile error would have
> > had plenty of opportunity in almost 14 years to surface.
>
> Oh, interesting. I didn't even know that the Windows runners had GCC in
> their PATH, and thus I didn't expect it to use that compiler at all. On
> GitLab for example we can see that it uses the MSVC compiler as I did
> expect [1]:
>
> Activating VS 17.10.2
> C compiler for the host machine: cl (msvc 19.40.33811 "Microsoft (R) C/C++ Optimizing Compiler Version 19.40.33811 for x64")
> C linker for the host machine: link link 14.40.33811.0
>
> But you're right, on GitHub that's not the case:
>
> C compiler for the host machine: gcc (gcc 12.2.0 "gcc (x86_64-posix-seh-rev2, Built by MinGW-W64 project) 12.2.0")
> C linker for the host machine: gcc ld.bfd 2.39
>
> We can easily fix that by passing the `--vsenv` flag to Meson. I'll send
> a patch soonish.
>
> Patrick
>
> [1]: https://gitlab.com/gitlab-org/git/-/jobs/9324989037#L95
> [2]: https://github.com/git/git/actions/runs/13686408338/job/38270746786#step:5:15
Please do not invest more time on the Visual Studio support via Meson. No
contributor will use this, and I want to stop spending my time on this.
The user experience of configuring a Visual Studio build via Meson is just
too weak compared to the ease of CMake-based builds, and while not many
Visual Studio users are familiar with CMake, even dramatically less will
even so much as know about Meson.
I plan on dropping all pretense that Git supports Visual Studio-based
contributions soon after v2.49.0 comes out, e.g. by deleting the CMake
definition and also deleting whatever Meson-specific stuff I can get away
deleting in Git for Windows. It was not worth the time I invested.
Ciao,
Johannes
| } | ||
| return pw; | ||
| } | ||
|
|
There was a problem hiding this comment.
On the Git mailing list, Junio C Hamano wrote (reply to this):
"Johannes Schindelin via GitGitGadget" <[email protected]>
writes:
> In other words, contrary to my expectations, the `win+Meson` job is
> ill-equipped to replace the `win build` job because it exercises a
> completely different tool version/compiler flags vector than what Git
> for Windows needs.
It is apparent that meson support is a new procedure to build our
codebase that is untested and unproven on Windows at all, given that
among all people who may have stake in Windows you are discovering
problems in it this late in the cycle. Nobody knows what other
breakages, other than something obvious and easy to catch like "ah,
compiler refuses to go further", are lurking under the radar.
I would be reluctant to trust the build artifact out of meson-based
build on Windows after seeing your report, especially the above
part.
A reasonable alternative may be to declare that meson-based build is
not ready yet at this point, and possibly disable win+Meson jobs to
punt and divert our engineering resources elsewhere in the meantime.
For a new thing, having an uneven support depending on the platform
early in the evolution is not unusual or to be ashamed of.
> Nevertheless, there is currently this huge push, including breaking
> changes after -rc1 and all, for switching to Meson. Therefore, we need
> to make it work, somehow, even in Git for Windows' SDK, hence this
> patch, at this point in time.
As I said earlier already, I do not mind turning the type of this
pointer, which is only used to read from a struct member, like this
patch does. It is the right thing to do, so I'll apply.
But I personally would not be comfortable with the product built
with "completely different tool version/compiler flags vector than
what G4W needs", even the compilation passes with just this small
change. If I were using Windows, that is.
> Signed-off-by: Johannes Schindelin <[email protected]>
> ---
> ident.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
Thanks, will apply.
> diff --git a/ident.c b/ident.c
> index caf41fb2a98..967895d8850 100644
> --- a/ident.c
> +++ b/ident.c
> @@ -59,7 +59,7 @@ static struct passwd *xgetpwuid_self(int *is_bogus)
>
> static void copy_gecos(const struct passwd *w, struct strbuf *name)
> {
> - char *src;
> + const char *src;
>
> /* Traditionally GECOS field had office phone numbers etc, separated
> * with commas. Also & stands for capitalized form of the login name.There was a problem hiding this comment.
On the Git mailing list, Patrick Steinhardt wrote (reply to this):
On Thu, Mar 06, 2025 at 08:33:43AM -0800, Junio C Hamano wrote:
> "Johannes Schindelin via GitGitGadget" <[email protected]>
> writes:
>
> > In other words, contrary to my expectations, the `win+Meson` job is
> > ill-equipped to replace the `win build` job because it exercises a
> > completely different tool version/compiler flags vector than what Git
> > for Windows needs.
>
> It is apparent that meson support is a new procedure to build our
> codebase that is untested and unproven on Windows at all, given that
> among all people who may have stake in Windows you are discovering
> problems in it this late in the cycle. Nobody knows what other
> breakages, other than something obvious and easy to catch like "ah,
> compiler refuses to go further", are lurking under the radar.
>
> I would be reluctant to trust the build artifact out of meson-based
> build on Windows after seeing your report, especially the above
> part.
>
> A reasonable alternative may be to declare that meson-based build is
> not ready yet at this point, and possibly disable win+Meson jobs to
> punt and divert our engineering resources elsewhere in the meantime.
> For a new thing, having an uneven support depending on the platform
> early in the evolution is not unusual or to be ashamed of.
I think it would be a bit sad to disable those jobs. They build and pass
the test suite alright in Git itself, even though they fail downstream
in Git for Windows. They help me quite a bit to ensure that I don't
regress anything that already is working while I'm iterating on the
current set of features. So in the end, I view them more as testing more
variants of Windows than replacing what we currently have, similar to
how we test Git on different Linux distributions.
I have said before that I'm very willing to help to figure out any
issues, regardless of which platform, and I stand by that statement --
if you see anything that is broken in this context and report the issue
to me I'll jump on it immediately.
> > Nevertheless, there is currently this huge push, including breaking
> > changes after -rc1 and all, for switching to Meson. Therefore, we need
> > to make it work, somehow, even in Git for Windows' SDK, hence this
> > patch, at this point in time.
>
> As I said earlier already, I do not mind turning the type of this
> pointer, which is only used to read from a struct member, like this
> patch does. It is the right thing to do, so I'll apply.
>
> But I personally would not be comfortable with the product built
> with "completely different tool version/compiler flags vector than
> what G4W needs", even the compilation passes with just this small
> change. If I were using Windows, that is.
That's completely fair. The CI job we have isn't meant to verify that we
have a G4W-compatible distribution falling out of it, it merely verifies
that we can build and pass tests in such a "standalone" (that is,
without the SDK) configuration. We might eventually want to introduce
another job that _does_ use the SDK with Meson, as well, but I didn't
yet see a need for that until now.
PatrickThere was a problem hiding this comment.
On the Git mailing list, Junio C Hamano wrote (reply to this):
Patrick Steinhardt <[email protected]> writes:
> I think it would be a bit sad to disable those jobs. They build and pass
> the test suite alright in Git itself, even though they fail downstream
> in Git for Windows. They help me quite a bit to ensure that I don't
> regress anything that already is working while I'm iterating on the
> current set of features. So in the end, I view them more as testing more
> variants of Windows than replacing what we currently have, similar to
> how we test Git on different Linux distributions.
Hmph, but compared to Linux or macOS platforms, do developers on
Windows (and users of Git on Windows, including but not limited to
users of Git-for-windows) benefit from having the code base to be
tested on "more variants of Windows", or would it be more noise that
they need to go visit the failing CI and spend time to triage if the
breakage is something they should worry about?
The above is more or less a rhetorical question. I think by now
everybody knows I do not like monoculture, and if we had infinite
engineering resources, I would think it would be healthy to have
more than one prominent and competing Windows port of Git (no, I
know about Cygwin, but I hear that the platform is POSIXy enough, so
I do not exactly consider Git running on Cygwin qualifies as "a
Windows port"). But we do not seem to live in such a world.
> I have said before that I'm very willing to help to figure out any
> issues, regardless of which platform, and I stand by that
> statement -- if you see anything that is broken in this context
> and report the issue to me I'll jump on it immediately.
It's ultimately up to Windows folks to take you up on the offer.
>> > Nevertheless, there is currently this huge push, including breaking
>> > changes after -rc1 and all, for switching to Meson. Therefore, we need
>> > to make it work, somehow, even in Git for Windows' SDK, hence this
>> > patch, at this point in time.
>> ...
> That's completely fair. The CI job we have isn't meant to verify that we
> have a G4W-compatible distribution falling out of it, it merely verifies
> that we can build and pass tests in such a "standalone" (that is,
> without the SDK) configuration. We might eventually want to introduce
> another job that _does_ use the SDK with Meson, as well, but I didn't
> yet see a need for that until now.
Knowing that it is the most widely used platform, I naturally and
naïvely was expecting and hoping that there are folks other than
Dscho who have enough bandwidth and inclination to help in this
area, but so far, having a set of jobs in CI that use a toolchain
that is different from what G4W uses (as expected) did not quite
help X-<. Nobody noticed it until the last minute.
Which made me say that we do not seem to live in such a world, which
in turn makes me accept that putting all Windows eggs in a single
basket and watch it closely may be a reasonable alternative when it
comes to Windows [*], than hoping that diverse set of different
builds eventually help flourishing Windows developer community.
I dunno.
[Footnote]
* Yes, I admit that it may be another way to say that I do not care
the particular platform deeply enough.
|
This patch series was integrated into seen via git@9530b9c. |
|
This patch series was integrated into next via git@7583a8b. |
|
This patch series was integrated into seen via git@a36e024. |
|
This patch series was integrated into master via git@a36e024. |
|
This patch series was integrated into next via git@4cd3354. |
|
Closed via a36e024. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
I needed many patches to make Git for Windows v2.49.0-rc0 compile and run the many, many CI jobs successfully. These here patches even apply to upstream Git. (Technically, the Meson sorting patch is not required to compile, but it was the fall-out from many required adjustments to make the Meson jobs happy.)
Changes since v1:
pw_gecosfield triggers, and augmented the commit message accordingly.cc: Patrick Steinhardt [email protected]