[PATCH] dispatch: fix so that --help more reliably shows command help (issue4240)

[Simon King]

On Thu, Jul 10, 2014 at 10:17 AM, Prabhu GS <pprabhugs at gmail.com> wrote:
> On Wed, Jul 9, 2014 at 7:53 AM, Augie Fackler <raf at durin42.com> wrote:
>>
>> On Tue, Jul 01, 2014 at 10:35:55PM +0530, Prabhu Gnana Sundar wrote:
>> > # HG changeset patch
>> > # User Prabhu Gnana Sundar <pprabhugs at gmail.com>
>> > # Date 1404234214 -19800
>> > #      Tue Jul 01 22:33:34 2014 +0530
>> > # Node ID 43e4cd31fae70005b4680ccb53b51f46b962e5b6
>> > # Parent  99db956b88ab699644c99095fecadbc4c83adbfc
>> > dispatch: fix so that --help more reliably shows command help
>> > (issue4240)
>>
>> After running tests, I'm dropping this. Please take a look at the
>> output of 'make test-execute-bit.t' (among many other breakages).
>>
>
> Hi all,
>
> I tried to see what went wrong. So i just started off by looking at what
> failed in test-cat.t test.
>
> When I ran the test-cat.t test, I could see the below snippet,
>
> <partial snippet>
>
> --- /home/prabhugs/projects/source/hg/tests/test-cat.t
> +++ /home/prabhugs/projects/source/hg/tests/test-cat.t.err
> @@ -17,7 +17,7 @@
>    $ hg cat -r 0 b
>    0
>    $ hg cat -r 1 a
> -  a: no such file in rev 7040230c159c
> +  a: no such file in rev ecd15c2a36c4
>    [1]
>
> </partial snippet>
>
> I could not really understand why the rev number of the commit changed ? The
> test-cat.t test just makes a commit and tries to cat the file from a
> revision where the file itself is non-existent.
>
> I could see that the commit revision number changes just with the
> mercurial/cmdutil.py patch.
> If I revert the mercurial/cmdutil.py patch and test it, the revision number
> is not changed and the test passes.
>
>
> Please help me with a few pointers...
>

The tests often use a default .hgrc file that adds "-d '0 0'" to
various commands so that checkins always appear to have the same date
(so that the hash will be stable). See:

  http://selenic.com/hg/file/3b97a93dc137/tests/run-tests.py#l645

Is it possible that your change breaks [defaults] operation for
commands that are aliased (eg. ci instead of checkin)?

Simon