RE: [BUGS] Git v2.51.2 on NonStop

[<rsbecker@nexbridge.com>]

On October 29, 2025 7:18 PM, SZEDER Gábor wrote:
>On Wed, Oct 29, 2025 at 06:41:27PM -0400, rsbecker@nexbridge.com wrote:
>> On October 29, 2025 6:28 PM, SZEDER Gábor wrote:
>> >On Tue, Oct 28, 2025 at 01:40:42PM -0400, rsbecker@nexbridge.com wrote:
>> >> I have found new defects on 2.51.2 that were not present in 2.51.1
>> >> when building
>> >with OpenSSL 3.5 (probably unrelated).
>> >>
>> >> Many failures in t7900 resulting from the use of test_subcommand !
>> >> as seen
>> >below. This is run in bash 5.0.18:
>> >>
>> >> + test_subcommand git gc --quiet --no-detach
>> >> + --skip-foreground-tasks 0< run-no-auto.txt
>> >> {"event":"child_start","sid":"20251028T172435.690328Z-H36b1c598-P7b
>> >> 000
>> >> 02d","thread":"main","time":"2025-10-28T17:24:35.879833Z","file":"/
>> >> hom
>> >> e/jenkinsbuild/.jenkins/workspace/Git_Pipeline/run-command.c","line
>> >> ":7
>> >> 40,"child_id":2,"child_class":"?","use_shell":false,"argv":["git","gc"
>> >> ,"--quiet","--no-detach","--skip-foreground-tasks"]}
>> >> + test_subcommand ! git gc --auto --quiet --no-detach
>> >> + --skip-foreground-tasks 0< run-auto.txt
>> >> test_subcommand[5]: test: argument expected
>> >
>> >This is a strange trace output, as it includes file descriptor
>> >redirections, but lacks the trace of commands executed in
>> >test_subcommands.  This trace doesn't look like it's coming from Bash.
>>
>> The test_subcommand[5]: test: argument expect is coming from bash. I
>> cannot see how it is coming from anywhere else. The test was run with
>> -x but snipped. Reference to test_subcommand ! git... fails exactly the same way.
>Is there something else I can provide?
>
>The -x trace from Bash doesn't include file descriptor redirections, but includes the
>commands executed in test_subcommand, and the number of '+' characters at the
>beginning of each trace line indicates the number of subshells, so when the test is
>really run by Bash the trace from a test_subcommand invocation looks like this:
>
>  ++ test_subcommand git gc --quiet --no-detach --skip-foreground-tasks
>  ++ local negate=
>  ++ test git = '!'
>  +++ printf '"%s",' git gc --quiet --no-detach --skip-foreground-tasks
>  ++ local 'expr="git","gc","--quiet","--no-detach","--skip-foreground-tasks",'
>  ++ expr='"git","gc","--quiet","--no-detach","--skip-foreground-tasks"'
>  ++ test -n ''
>  ++ grep '\["git","gc","--quiet","--no-detach","--skip-foreground-tasks"\]'
>  {"event":"child_start","sid":"20251029T230304.507599Z-H3fb05a16-
>P001a33ae","thread":"main","time":"2025-10-29T23:03:04.512913Z","file":"run-
>command.c","line":740,"child_id":2,"child_class":"?","use_shell":false,"argv":["git",
>"gc","--quiet","--no-detach","--skip-foreground-tasks"]}
>  ++ test_subcommand '!' git gc --auto --quiet --no-detach --skip-foreground-tasks
>  ++ local negate=
>
>Since the trace you showed doesn't look like this I still doubt that it came from Bash.

What appears to be happening is that the Make environment is only using SHELL=/bin/bash
for the outer processing but not the inner #!/bin/sh of t7900. The system is using /bin/sh
as specified, which uses ksh, not bash, which is the trace above. When I run the individual
tests with bash, the error reported goes away. The problem is, with my version of Gnu Make,
4.1.2, the SHELL variable is only being replaced for the command processing of each
recipe. Once the system loader sees the shebang of #!/bin/sh, /bin/sh is used as requested,
and fails out. This means that I have to remember to manually run each test that fails with
bash instead of the default. It is frustrating and now adds hours to my manual evaluation
of the CI/CD results.

The trace above is from sh, not bash because of this.

With Frustration,
Randall