Re: [PATCH 3/3] t7900: make macOS-specific test work on Windows

From: Eric Sunshine <sunshine@sunshineco.com>
Date: Nov 28, 2020, 05:55 UTC
Message-ID: <CAPig+cRbga9ksi0JedU=U1zbLMY26xLtB7sBDJ5M5=TT1HsX1A@mail.gmail.com>
In-Reply-To: <87o8jikfh7.fsf@evledraar.gmail.com>

On Fri, Nov 27, 2020 at 10:05 AM Ævar Arnfjörð Bjarmason <avarab@gmail.com> wrote:

Show 12 quoted lines

> On Fri, Nov 27 2020, Eric Sunshine wrote:
> > +test_expect_success 'start and stop macOS maintenance' '
> > +     uid=$(test-tool getuid) &&
>
> This seems equally portable, and means your 2/3 isn't needed, no?
>     +   uid=FAKE_UID &&
>     +   echo \$* | perl -pe "s[(?<= gui/)-?[0-9]+][$uid]g" >>args
> I.e. the context here is that the test is already hardcoding an
> assumption about "gui/%d" (per code in gc.c). It seems more robust & to
> the point of the test to not care about the specific UID number that
> comes back, since we're really testing whether we invoke our own code,
> not platform getuid() sanity.

Good idea. That's a reasonable viewpoint to take, and makes the series simpler.