[PATCH 3 of 3] test-progress: disable mocking-time tests on chg
Pierre-Yves David
pierre-yves.david at ens-lyon.org
Thu Apr 14 15:23:08 EDT 2016
On 04/13/2016 03:41 AM, Jun Wu wrote:
> On 04/13/2016 07:42 AM, Pierre-Yves David wrote:
>> Actually, not entirely. If fixing all extensions immediatly is too
>> much a
>> trouble we could have the devel warning disabled, either by default
>> or for
>> specific tests. Fixing them iteratively after than.
>
> How about other issues?
>
> 1. Hacking stdlib is ugly
Yes it is. But we don't have a better option, do we? Given that the
monkey patch will be temporary and only in the case where devel-warning
are enabled, I think this is okay. We do some awful stuff in core from
time to time when we don't have better option.
> 2. chg will only display devel-warn in extensions.loadall once (i.e. a
> second
> run won't display these warnings)
This is fine, it will be caught by the non-chg runs. Especially the
tests are run with the devel-warning on and I expect most people to see
them there.
> 3. Implementing the devel-warn itself require large refactoring if
> "import
> chgserver" in extensions.py is not allowed.
Can you elaborate about this refactoring? From my understanding, the
list of "watched" config and env were to be moved into extensions and
that is it. Anything else is required?
>
>
>> This would avoid implementing the feature twice, once in the script, and
>> once for the markers. It also increase the chance that we actually
>> end up
>> with a devel warning.
>
> There are too many tests related so only disabling by default makes
> sense.
> If it is disabled by default then we lose the meaning of having it:
> prevent
> new code from breaking chg.
How many test actually breaks? How many core extensions are currently
broken?
Having a devel warning will allow us to slowly fixing all the issues,
when they will all be fixed (or blacklisted) we can just get the benefit
by flipping a small switch.
Cheers,
--
Pierre-Yves David
More information about the Mercurial-devel
mailing list