Re: [PATCH] t3700-add.sh: Avoid filename collission between tests which could lead to test failure

From: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
Date: Oct 10, 2016, 17:09 UTC
Message-ID: <D7AF680C-1BED-431C-980E-5895B3848815@apple.com>
In-Reply-To: <xmqqlgxwgnmu.fsf@gitster.mtv.corp.google.com>

Show 12 quoted lines

> On Oct 10, 2016, at 09:44, Junio C Hamano <gitster@pobox.com> wrote:
> 
> Jeremy Huddleston Sequoia <jeremyhu@apple.com> writes:
> 
>> Regressed-in: 610d55af0f082f6b866dc858e144c03d8ed4424c
>> Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
>> CC: Thomas Gummerer <t.gummerer@gmail.com>
>> CC: Junio C Hamano <gitster@pobox.com>
> 
> This is also under-explained.  Was the test broken for everybody and
> you are the first to report, or was your environment somehow different
> that the existing problem was revealed there first?

There is nothing obviously different about my environment.  I'd expect this test to fail for everyone that runs the entire test because xfoo3 is not cleaned up after this earlier test:

test_expect_success \
        'git update-index --add: Test that executable bit is not used...' \
        'git config core.filemode 0 &&
         test_ln_s_add xfoo2 xfoo3 &&   # runs git update-index --add
         test_mode_in_index 120000 xfoo3'

If anyone ran the test by itself or otherwise skipped the 'git update-index --add: Test that executable bit is not used...' test, they would not encounter the issue.  However, the normal case is to run all of the tests, so I would expect everyone to be tripping over this issue.  If they're not, then that itself might be indicative of another bug because clearly that previous test just added xfoo3 as a symlink.

--Jeremy