Re: [PATCH v4 0/5] meson: add corresponding target for Makefile's hdr-check

From: Junio C Hamano <gitster@pobox.com>
Date: Apr 21, 2025, 07:48 UTC
Message-ID: <xmqqh62i6jli.fsf@gitster.g>
In-Reply-To: <20250420-505-wire-up-sparse-via-meson-v4-0-66e14134e822@gmail.com>

Karthik Nayak <karthik.188@gmail.com> writes:

Show 27 quoted lines

> This is based on master 9d22ac5122 (The third batch, 2025-04-07) with
> 'es/meson-build-skip-coccinelle' merged in.
>
> ---
> Changes in v4:
> - Rename headers to headers_to_check, since these headers are only used
>   for static analysis.
> - Added a commit to rename third_party_sources -> third_party_excludes
>   and remove a duplicate.
> - Fix a typo 'gcrpyt' -> 'gcrypt'
> - Remove 'generated_headers', since we use 'git ls-files' and that would
>   already ignore files within '.gitignore'.
> - Link to v3: https://lore.kernel.org/r/20250414-505-wire-up-sparse-via-meson-v3-0-edc6e7f26745@gmail.com
>
> Changes in v3:
> - Some renames:
>   - headers_generated -> generated_headers
>   - meson -> Meson
>   - headers-check -> check-headers
>   - headers_check_exclude -> exclude_from_check_headers
> - Rewrite 'headers_check_exclude' to also contain dirs so we can skip
>   listing individual header files.
> - Move 'xdiff/*' to 'third_party_sources' and cleanup
>   'exclude_from_check_headers'.
> - Use 'echo' instead of 'echo -n'.
> - Use `fs.replace_suffix` instead of `str.replace`.
> - Link to v2: https://lore.kernel.org/r/20250410-505-wire-up-sparse-via-meson-v2-0-acb45cc8a2e5@gmail.com

Like the previous round, this round also seems to break linux.meson job at GitHub Actions CI when merged to 'seen'. It may be quite possible that it is caused by some semantic conflicts, and help to find where the merged result is wrong is very much appreciated.

For now, I've ejected the topic out of 'seen' again.

Failing CI run:

https://github.com/git/git/actions/runs/14563669225/job/40850047961

The same 'seen' without the merge of this topic:

https://github.com/git/git/actions/runs/14564707024/job/40852298435

Note that I expect win+Meson test (3) job to fail/time-out even without this topic. I do not know how the numbered test jobs for win+Meson part of the CI are sharded, but one of them never finishes in 'next' so this topic is unlikely to be the cause of the breakage.

Addendum.

The topic alone, rebuilt as specified in your cover letter, seems to be sufficient to make linux-meson job fail, without any interaction with other topics in 'seen'.

https://github.com/git/git/actions/runs/14566676782/job/40856998621

Note that we have unrelated breakages like pedantic (due to update to the fedora image) and sparse (due to deprecation of the ubuntu image) we fixed since the base commit you picked, so failures of these jobs are not all that interesting in this run. But the breakage of linux-meson job is unique to this topic.

Interestingly, either this topic alone or 'seen' with this topic I haven't seen the "meson setup + compile + test" sequence in my local testing, so the breakage seems to be isolated to the CI job.