Re: [PATCH v2] CI: select CC based on CC_PACKAGE (again)

["Ævar Arnfjörð Bjarmason" <avarab@gmail.com>]

On Thu, Apr 21 2022, Junio C Hamano wrote:

> Ævar Arnfjörð Bjarmason  <avarab@gmail.com> writes:
>
>> Fix a regression in 707d2f2fe86 (CI: use "$runs_on_pool", not
>> "$jobname" to select packages & config, 2021-11-23).
>>
>> In that commit I changed CC=gcc from CC=gcc-9, but on OSX the "gcc" in
>> $PATH points to clang, we need to use gcc-9 instead. Likewise for the
>> linux-gcc job CC=gcc-8 was changed to the implicit CC=gcc, which would
>> select GCC 9.4.0 instead of GCC 8.4.0.
>
> Thanks for diagnosing how things were broken.
>
>> On Thu, Apr 21 2022, Phillip Wood wrote:
>>
>>> CC is set in .github/workflows/main.yaml for the ubuntu and macos jobs
>>> so I think they will not fallback to using CC_PACKAGE and therefore
>>> not pick up the correct compiler.
>> ...
>> diff --git a/.github/workflows/main.yml b/.github/workflows/main.yml
>> index c35200defb9..f12819a00d7 100644
>> --- a/.github/workflows/main.yml
>> +++ b/.github/workflows/main.yml
>> @@ -236,7 +236,6 @@ jobs:
>>            - jobname: linux-TEST-vars
>>              cc: gcc
>>              os: ubuntu
>> -            cc_package: gcc-8
>>              pool: ubuntu-latest
>>            - jobname: osx-clang
>>              cc: clang
>> diff --git a/ci/lib.sh b/ci/lib.sh
>> index cbc2f8f1caa..86e37da9bc5 100755
>> --- a/ci/lib.sh
>> +++ b/ci/lib.sh
>> @@ -122,7 +122,7 @@ then
>>  	test macos != "$CI_OS_NAME" || CI_OS_NAME=osx
>>  	CI_REPO_SLUG="$GITHUB_REPOSITORY"
>>  	CI_JOB_ID="$GITHUB_RUN_ID"
>> -	CC="${CC:-gcc}"
>> +	CC="${CC_PACKAGE:-${CC:-gcc}}"
>>  	DONT_SKIP_TAGS=t
>>  
>>  	cache_dir="$HOME/none"
>
> OK, so we favor CC_PACKAGE (from the matrix.vector.cc_package) if
> set, and then cc (again, from the matrix.vector.cc) if set, and then
> finally use "gcc" as a fallback.  In the osx-gcc job, cc_package is
> set to gcc-9 while in the osx-clang, cc is gcc that confusingly calls
> for clang.  That sounds like it would do the right thing for two
> macs.

Yes.

> For other jobs with different settings for cc and cc_package, does
> this have any effect?  I do not think I saw any mention in the
> proposed log message.
>
> 		vector.cc	vector.cc_package	old	new
> linux-clang	clang		-			clang	clang
> linux-sha256	clang		-			clang	clang
> linux-gcc	gcc		gcc-8			gcc	gcc-8
> osx-clang	clang		-			clang	clang
> osx-gcc		gcc		gcc-9			clang	gcc-9
> linux-gcc-default gcc		-			gcc	gcc
>
> So, linux-gcc job used to use whichever "gcc" the platform gave us,
> but now it explicitly asks for gcc-8, which may or may not be
> different from what linux-gcc-default uses, and there is no other
> difference by this change.  We may get a better test coverage (if
> the default gcc is not gcc-8) or no improvement (if the default is
> gcc-8), so it is a strict improvement worth recording as an intended
> side effect in the proposed log message to help future developers.
>
> Other than that, looks good to me.

I'm happy to rephrase it however you'd like, but I'm a bit confused by
the "saw any mention in the proposed log message". I'm fairly sure
paragraph 2 onwards covers this, i.e. how linux-gcc's behavior is
changed (as it also regressed).

What I suppose is left undiscussed is that jobs that don't define
CC_PACKAGE at all won't be impacted, is that what you wanted to be
explicitly mentioned?