[merge][0.16][#111968] error in _run_hooks when pulling or pushing in a branch bound to a remote branch

[Ian Clatworthy]

>
>> +        # XXX: It looks like there are some orders for generating tests
>> that
>> +        # fail as of 20070504 - maybe because of import order
>> dependencies.
>> +        # So unfortunately this will rarely intermittently fail at the
>> moment.
>> +        # -- mbp 20070504
>>
>> Didn't you fix that? I don't really like having an XXX: that we can't
>> prove, but if you feel it is helpful for people encountering the bug,
>> I'm okay with leaving it in.
>
> I fixed at least one of them but I suspect there might be more.  The
> thing to do is to run selftest --randomize repeatedly until it fails,
> then fix the failure - or conclude after many runs that there are no
> more such bugs.
>
> I dislike having tests with random components, but we do need to test
> the option and I suppose this is reasonably confined, since it doesn't
> actually run the tests.
>
Just a reminder that the randomize option supports a SEED parameter so 
that others can run with exactly the same "random" order. The SEED for 
any given run is displayed in the output. When reporting bugs found by 
using randomize, I think it's a good policy to include the SEED in the 
bug report.


Ian C.