Re: What's cooking in git.git (Apr 2011, #02; Wed, 6)

[Jens Lehmann <Jens.Lehmann@web.de>]

Am 11.04.2011 11:00, schrieb Ævar Arnfjörð Bjarmason:
> On Mon, Apr 11, 2011 at 00:55, Jens Lehmann <Jens.Lehmann@web.de> wrote:
>> Am 10.04.2011 20:52, schrieb Ævar Arnfjörð Bjarmason:
>>> On Wed, Apr 6, 2011 at 22:04, Junio C Hamano <gitster@pobox.com> wrote:
>>>
>>>> * jl/submodule-fetch-on-demand (2011-03-06) 7 commits
>>>>  (merged to 'next' on 2011-03-20 at a5e452d)
>>>>  + fetch/pull: Describe --recurse-submodule restrictions in the BUGS section
>>>>  + submodule update: Don't fetch when the submodule commit is already present
>>>>  + fetch/pull: Don't recurse into a submodule when commits are already present
>>>>  + Submodules: Add 'on-demand' value for the 'fetchRecurseSubmodule' option
>>>>  + config: teach the fetch.recurseSubmodules option the 'on-demand' value
>>>>  + fetch/pull: Add the 'on-demand' value to the --recurse-submodules option
>>>>  + fetch/pull: recurse into submodules when necessary
>>>
>>> This series needs to be fixed to not fail tests with GETTEXT_POISON=YesPlease.
>>
>> Maybe something like this? Junio, do you want me to squash the changes in
>> the patch series and resend it or should I send a new commit?

Ok, as that series already hit master I'll hack up a new commit.

> Sorry about the short mail yesterday, I had to take care of something.

Same here :-)

> Anyway, that looks like it'll work, but the preferred way of doing it
> is like it's done at the bottom of t3700-add.sh. I.e.:
> 
> test_expect_success 'git add --dry-run --ignore-missing of
> non-existing file' '
>     test_must_fail git add --dry-run --ignore-missing track-this
> ignored-file >actual.out 2>actual.err
> '
> 
> test_expect_success C_LOCALE_OUTPUT 'git add --dry-run
> --ignore-missing of non-existing file output' '
>     test_cmp expect.out actual.out &&
>     test_cmp expect.err actual.err
> '
> 
> That is, split up each test into a test that tests the functionality
> (i.e. exit codes) and one that tests the output.
> 
> That way we make sure that only things dependent on C locale output
> are skipped under non-C locales.

Thanks for explaining, that makes sense. Jonathan mentioned some time ago
that t5526 would benefit from testing the functionality rather than the
output like it does now, so rewriting that test is on my maintenance list
anyway. I'll have to find a locale independent way to check if the
submodule paths are printed correctly, but I'll see if I can chop off the
language specific parts of the output before doing the compare.

> Sometimes that's hard, in which case it's fine to just skip the entire thing.

That shouldn't be necessary for t5526, I'll tackle that when I find some
time. The question is: Do we need a commit now adding the prerequisite as
a band aid fix to restore the ability to successfully run the test suite
with GETTEXT_POISON=YesPlease, or do we have some time so I can rewrite
t5526 properly to not depend on the language anymore?