[Bug other/111368] New: contrib/test_summary should check to ensure the git branch being put in the email matches the branch actually tested

[Bug other/111368] New: contrib/test_summary should check to ensure the git branch being put in the email matches the branch actually tested

[egallager at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111368

            Bug ID: 111368
           Summary: contrib/test_summary should check to ensure the git
                    branch being put in the email matches the branch
                    actually tested
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: other
          Assignee: unassigned at gcc dot gnu.org
          Reporter: egallager at gcc dot gnu.org
  Target Milestone: ---

contrib/gcc_update will put the git branch being updated into gcc/REVISION.
That's cool, except sometimes I'll change branches and forget to run
contrib/gcc_update, meaning I'll have a stale gcc/REVISION left over. The
branch name put in there will get used in other places, as well. Meanwhile,
contrib/test_summary will put the branch name that it finds into the email
subject line that it generates. Take a look at, say:
https://gcc.gnu.org/pipermail/gcc-testresults/2023-September/795648.html
It says: "[master r14-3825-g0d50fac]"
However, this wasn't actually the branch that I had tested in that particular
set of test results; it should actually be users/me/autotools-tinkering in that
case. The contrib/test_summary script should test the git branch that it's
printing against the git branch that was actually tested to help guard against
errors like this.

[Bug other/111368] contrib/test_summary should check to ensure the git branch being put in the email matches the branch actually tested

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111368

--- Comment #1 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to Eric Gallager from comment #0)
> The contrib/test_summary script
> should test the git branch that it's printing against the git branch that
> was actually tested to help guard against errors like this.

How can it know the git branch that was actually tested?

The test_summary script just looks at the contents of the source tree and the
build directory, specifically at the LAST_UPDATED file and the output of
config.status. If the GCC build that was tested lies about its version, then I
don't think it should be the test_summary script's job to detect that.

[Bug other/111368] contrib/test_summary should check to ensure the git branch being put in the email matches the branch actually tested

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111368

--- Comment #2 from Jonathan Wakely <redi at gcc dot gnu.org> ---
N.B. test_summary doesn't print that "[master r14-3825-g0d50fac]" string
itself, it just scrapes it from the logs, and the string in the logs comes
straight from the xgcc. xg++ etc. binaries.

I don't think it's the test_summary script's job to correct bad info in the
logs.

[Bug other/111368] contrib/test_summary should check to ensure the git branch being put in the email matches the branch actually tested

[egallager at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111368

--- Comment #3 from Eric Gallager <egallager at gcc dot gnu.org> ---
(In reply to Jonathan Wakely from comment #2)
> N.B. test_summary doesn't print that "[master r14-3825-g0d50fac]" string
> itself, it just scrapes it from the logs, and the string in the logs comes
> straight from the xgcc. xg++ etc. binaries.
> 
> I don't think it's the test_summary script's job to correct bad info in the
> logs.

Well ok, maybe the check could be performed earlier, then, say build time or
configure time?