Re: [PATCH] name-rev: use generation numbers if available

From: Junio C Hamano <gitster@pobox.com>
Date: Mar 1, 2022, 07:33 UTC
Message-ID: <xmqqfso2t9cu.fsf@gitster.g>
In-Reply-To: <CA+P7+xoECs-rXb4vpRrw40Q-oRvfu97kMig9zu0rEE6KagAyiw@mail.gmail.com>

Jacob Keller <jacob.keller@gmail.com> writes:

Show 29 quoted lines

> On Mon, Feb 28, 2022 at 6:36 PM Junio C Hamano <gitster@pobox.com> wrote:
>>
>> Jacob Keller <jacob.e.keller@intel.com> writes:
>>
>> > +test_expect_success 'name-rev without commitGraph does not handle non-monotonic timestamps' '
>> > +     test_config -C non-monotonic core.commitGraph false &&
>> > +     (
>> > +             cd non-monotonic &&
>> > +
>> > +             rm -rf .git/info/commit-graph* &&
>> > +
>> > +             echo "main~3 undefined" >expect &&
>> > +             git name-rev --tags main~3 >actual &&
>> > +
>> > +             test_cmp expect actual
>> > +     )
>> > +'
>>
>> I doubt it is wise to "test" that a program does _not_ produce a
>> correct output, or even worse, it produces a particular wrong
>> output.  This test, for example, casts in stone that any future
>> optimization that does not depend on the commit-graph is forever
>> prohibited.
>>
>> Just dropping the test would be fine, I would think.
>
> Stolee mentioned it. We could also convert it to a
> "test_expect_failure" with the expected output too... But that makes
> it look like something we'll fix

Neither sounds like a good idea anyway. What we care most is with commit graph, the algorithm will not be fooled by skewed timestamps.