From: Junio C Hamano <gitster@pobox.com>
To: Siddharth Shrimali <r.siddharth.shrimali@gmail.com>
Cc: git@vger.kernel.org, peff@peff.net
Subject: Re: [PATCH] t3700: avoid suppressing git's exit code
Date: Fri, 27 Feb 2026 10:19:48 -0800 [thread overview]
Message-ID: <xmqqv7fiqcaj.fsf@gitster.g> (raw)
In-Reply-To: <20260227165143.70188-1-r.siddharth.shrimali@gmail.com> (Siddharth Shrimali's message of "Fri, 27 Feb 2026 22:21:43 +0530")
Siddharth Shrimali <r.siddharth.shrimali@gmail.com> writes:
> When piping the output of git ls-files into grep, the exit code of
> git ls-files is suppressed.
>
> Avoid this by redirecting the output of git ls-files to a file and
> then running grep on that file. This ensures that any crash in
> git ls-files will be caught by the test suite.
>
> Signed-off-by: Siddharth Shrimali <r.siddharth.shrimali@gmail.com>
> ---
> t/t3700-add.sh | 38 +++++++++++++++++++++++---------------
> 1 file changed, 23 insertions(+), 15 deletions(-)
A few things that I noticed.
* If "git foo | grep bar" was expecting to hide exit code from
"git" and check the output (e.g., "git diff --exit-code | grep foo"),
a mechanical conversion "git diff --exit-code >out && grep foo out"
would change the meaning of the test and break it. I did not check
if this patch has such an unintended breakage, though.
* Many of them do this:
> - git ls-files foo | grep foo
> + git ls-files foo >actual &&
> + grep foo <actual
or this
> - ! ( git ls-files foo1 | grep foo1 )
> + git ls-files foo1 >actual &&
> + ! grep foo1 actual
in which we might consider using "test_grep" (and "test_grep !")
to help the developer who wants to debug a breakage in ls-files
by highlighting what is unexpected in the output in their broken
version.
* A rewrite like this may want to be further broken down.
> - test $(git ls-files --stage | grep ^100644 | wc -l) -eq 0 &&
> + test $(git ls-files --stage >actual && grep ^100644 actual | wc -l) -eq 0 &&
If "ls-files --stage" segfaults, "grep | wc" would not run, $()
may exit with non-zero and turn into an empty string, but the
final error diagnosis would be something unfathonable like
test: -eq unary operator expected
test: missing argument after '0'
which would not help the person debugging the test very much.
next prev parent reply other threads:[~2026-02-27 18:19 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-02-27 16:51 [PATCH] t3700: avoid suppressing git's exit code Siddharth Shrimali
2026-02-27 18:19 ` Junio C Hamano [this message]
2026-02-28 7:00 ` [PATCH v2] t3700: avoid hidden failures and use test_grep helper Siddharth Shrimali
2026-03-02 21:24 ` Junio C Hamano
2026-02-28 8:12 ` [PATCH] t3700: avoid suppressing git's exit code Johannes Sixt
2026-02-28 10:15 ` Siddharth Shrimali
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=xmqqv7fiqcaj.fsf@gitster.g \
--to=gitster@pobox.com \
--cc=git@vger.kernel.org \
--cc=peff@peff.net \
--cc=r.siddharth.shrimali@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox