On 2025-06-07 10:38:12 [+0200], Jonathan Nieder wrote:

Hi Sebastian,

Hi Jonathan,

Sebastian Andrzej Siewior wrote:

Even after the fix from https://bugs.debian.org/1102106, git ftbfs on
s390x. Filing as a separate bug because this is a different failure.

The patch you took was the top-most in the fix series, you also need the few that were applied before that one. That last one depends on the previous changes. I took me a while to figure that out.

I collected them and verified that the package builds on amd64 and
s390x.

Interesting! Part of what is surprising to me is that the test
failure is different (mailsplit instead of min-batch-size) - I'm also
not seeing where in the proposed patches would affect this
functionality, as opposed to improving the code style (though
improving the code style is also of course nice :)). Do you have a
link to the log from when you built on s390x? Was it running tests?

I rebuilt the whole package so all tests were running
https://breakpoint.cc/log-uns-patched.txt

The other thing that surprised me was that experimental, which
includes all patches in "next", also continued to fail tests on 64-bit
big endian architectures, though now that I check again in the
experimental buildd logs that is yet another, different failure!
From [1]:

This is
https://lore.kernel.org/git/[email protected]

Sebastian

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On 2025-06-07 11:31:07 [+0200], Jonathan Nieder wrote:

New uploads coming later today. :)

I don't want to rush or anything but… If you want me to prepare an
unblock request before the upload happens or so, just say so ;)

For the future: in general if you prepare a "git format-patch" format
patch or merge request in Salsa against the packaging repo, that tends
to be very welcome, since it makes it easier to use your work as-is.
I'd also be happy if you'd like to join as a co-maintainer (and we can
use the packaging repo as our shared space for working).

merge requests in Salsa? I wasn't aware that the git package is there.
The package points only to git.or.cz.
If you are all alone and would like some support I could try to join.
I usually manage to keep myself busy :)

Sincerely,
Jonathan

Sebastian

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

From: [email protected]

Source: git
Version: 1:2.49.0-2
Severity: serious
Tags: ftbfs

Even after the fix from https://bugs.debian.org/1102106, git ftbfs on
s390x. Filing as a separate bug because this is a different failure.

From [1]:

| expecting success of 4014.200 '-c format.mboxrd format-patch':
| sp=" " &&
| cat >msg <<-INPUT_END &&
| mboxrd should escape the body
|
| From could trip up a loose mbox parser
| >From extra escape for reversibility
| >>From extra escape for reversibility 2
| from lower case not escaped
| Fromm bad speling not escaped
| From with leading space not escaped
|
| F
| From
| From$sp
| From $sp
| From $sp
| INPUT_END
|
| cat >expect <<-INPUT_END &&
| >From could trip up a loose mbox parser
| >>From extra escape for reversibility
| >>>From extra escape for reversibility 2
| from lower case not escaped
| Fromm bad speling not escaped
| From with leading space not escaped
|
| F
| From
| From
| From
| From
| INPUT_END
|
| C=$(git commit-tree HEAD^^{tree} -p HEAD <msg) &&
| git -c format.mboxrd format-patch --stdout -1 $C~1..$C >patch &&
| git format-patch --pretty=mboxrd --stdout -1 $C~1..$C >compat &&
| test_cmp patch compat &&
| git grep -h --no-index -A11 \
| "^>From could trip up a loose mbox parser" patch >actual &&
| test_cmp expect actual
|
| --- expect 2025-05-29 16:44:17.156655735 +0000
| +++ actual 2025-05-29 16:44:17.164655735 +0000
| @@ -1,12 +1 @@
| >From could trip up a loose mbox parser
| ->>From extra escape for reversibility
| ->>>From extra escape for reversibility 2
| -from lower case not escaped
| -Fromm bad speling not escaped
| - From with leading space not escaped
| -
| -F
| -From
| -From
| -From
| -From
| not ok 200 - -c format.mboxrd format-patch

[1] https://buildd.debian.org/status/fetch.php?pkg=git&arch=s390x&ver=1%3A2.49.0-2&stamp=1748537058&raw=0

--- SoupGate-Win32 v1.05
* Origin: you cannot sedate... all the things you hate (1:229/2)

From: [email protected]

Hi Sebastian,

Sebastian Andrzej Siewior wrote:

Even after the fix from https://bugs.debian.org/1102106, git ftbfs on
s390x. Filing as a separate bug because this is a different failure.

The patch you took was the top-most in the fix series, you also need the
few that were applied before that one. That last one depends on the
previous changes. I took me a while to figure that out.

I collected them and verified that the package builds on amd64 and
s390x.

Interesting! Part of what is surprising to me is that the test
failure is different (mailsplit instead of min-batch-size) - I'm also
not seeing where in the proposed patches would affect this
functionality, as opposed to improving the code style (though
improving the code style is also of course nice :)). Do you have a
link to the log from when you built on s390x? Was it running tests?

The other thing that surprised me was that experimental, which
includes all patches in "next", also continued to fail tests on 64-bit
big endian architectures, though now that I check again in the
experimental buildd logs that is yet another, different failure!
From [1]:

| expecting success of 0001.37 'bare & --separate-git-dir incompatible within worktree':
| test_when_finished "rm -rf bare.git linkwt seprepo" &&
| test_commit gumby &&
| git clone --bare . bare.git &&
| git -C bare.git worktree add --detach ../linkwt &&
| test_must_fail git -C linkwt init --separate-git-dir seprepo 2>err &&
| test_grep "incompatible" err
|
| BUG: read-cache.c:338: unsupported ce_mode: 24440200000
| Aborted
| not ok 37 - bare & --separate-git-dir incompatible within worktree

Looking closer.

Thanks,
Jonathan

[1] https://buildd.debian.org/status/fetch.php?pkg=git&arch=s390x&ver=1%3A2.50.0%7Erc0%2Bnext.20250528-1&stamp=1748537212&raw=0

--- SoupGate-Win32 v1.05
* Origin: you cannot sedate... all the things you hate (1:229/2)

From: [email protected]

Sebastian Andrzej Siewior wrote:

On 2025-06-07 10:38:12 [+0200], Jonathan Nieder wrote:

Sebastian Andrzej Siewior wrote:

The patch you took was the top-most in the fix series, you also need the >>> few that were applied before that one. That last one depends on the
previous changes. I took me a while to figure that out.

I collected them and verified that the package builds on amd64 and
s390x.

Interesting! Part of what is surprising to me is that the test
failure is different (mailsplit instead of min-batch-size) - I'm also
not seeing where in the proposed patches would affect this
functionality, as opposed to improving the code style (though
improving the code style is also of course nice :)). Do you have a
link to the log from when you built on s390x? Was it running tests?

I rebuilt the whole package so all tests were running
https://breakpoint.cc/log-uns-patched.txt

Okay, that indeed leans toward my having made a mistake while
backporting. Thanks, I'll take the whole series instead like you've
suggested.

The other thing that surprised me was that experimental, which
includes all patches in "next", also continued to fail tests on 64-bit
big endian architectures, though now that I check again in the
experimental buildd logs that is yet another, different failure!
From [1]:

This is
https://lore.kernel.org/git/[email protected]

Ah, very nice! Thanks much for this.

New uploads coming later today. :)

For the future: in general if you prepare a "git format-patch" format
patch or merge request in Salsa against the packaging repo, that tends
to be very welcome, since it makes it easier to use your work as-is.
I'd also be happy if you'd like to join as a co-maintainer (and we can
use the packaging repo as our shared space for working).

Sincerely,
Jonathan

--- SoupGate-Win32 v1.05
* Origin: you cannot sedate... all the things you hate (1:229/2)