Create MSan config for Chromium recipe module + add WebRTC MSan recipe.

Create MSan config for Chromium recipe module + add WebRTC MSan recipe.

Refactor the MSan settings into a proper recipe module config.
Add a FYI builder using it for WebRTC.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=292214

[kjellander_chromium, 2014-09-29 08:41:32]

kjellander@chromium.org changed reviewers:
+ phajdan.jr@chromium.org

[kjellander_chromium, 2014-09-29 08:41:48]

Patchset #1 (id:1) has been deleted

[kjellander_chromium, 2014-09-29 08:44:53]

kjellander@chromium.org changed reviewers:
+ earthdok@chromium.org

[kjellander_chromium, 2014-09-29 08:44:53]

I believe the additional GYP_DEFINES for the testers will not have any negative
effect. 

Let me know if you think the single entry "msan=1" should be separated for the
testers (but I think that would add unnecessary confusion).

[Paweł Hajdan Jr., 2014-09-29 08:45:50]

LGTM

[earthdok, 2014-09-29 13:25:08]

https://codereview.chromium.org/612753004/diff/20001/scripts/slave/recipe_mod...
File scripts/slave/recipe_modules/chromium/config.py (right):

https://codereview.chromium.org/612753004/diff/20001/scripts/slave/recipe_mod...
scripts/slave/recipe_modules/chromium/config.py:340:
gyp_defs['msan_track_origins'] = 0
msan_track_origins=0 should not be a part of the config. This option should be
chosen individually for each bot configuration as it has a significant impact on
performance. We already set a default value for it in common.gypi (currently 1).

https://codereview.chromium.org/612753004/diff/20001/scripts/slave/recipe_mod...
scripts/slave/recipe_modules/chromium/config.py:341:
gyp_defs['use_instrumented_libraries'] = 1
I'm also against adding instrumented libraries to the msan config, because
certain configurations do not need them. For example, we plan to have an MSan
bot for PDFium, which has no DSO dependencies.

Can you factor out those settings to a separate config 'instrumented_libraries',
and then apply it to the recipe using chromium_apply_config? Another benefit of
this arrangement is that it allows us to use instrumented libraries with other
sanitizers.

(Note that I don't actually know what chromium_apply_config does, just guessing
from the name.)

[kjellander_chromium, 2014-09-30 10:56:40]

earthdok: PTAL

https://codereview.chromium.org/612753004/diff/20001/scripts/slave/recipe_mod...
File scripts/slave/recipe_modules/chromium/config.py (right):

https://codereview.chromium.org/612753004/diff/20001/scripts/slave/recipe_mod...
scripts/slave/recipe_modules/chromium/config.py:340:
gyp_defs['msan_track_origins'] = 0
On 2014/09/29 13:25:08, earthdok wrote:
> msan_track_origins=0 should not be a part of the config. This option should be
> chosen individually for each bot configuration as it has a significant impact
on
> performance. We already set a default value for it in common.gypi (currently
1).

That makes sense, I left that variable in the existing config and used the
default for WebRTC in PS#2.

https://codereview.chromium.org/612753004/diff/20001/scripts/slave/recipe_mod...
scripts/slave/recipe_modules/chromium/config.py:341:
gyp_defs['use_instrumented_libraries'] = 1
On 2014/09/29 13:25:08, earthdok wrote:
> I'm also against adding instrumented libraries to the msan config, because
> certain configurations do not need them. For example, we plan to have an MSan
> bot for PDFium, which has no DSO dependencies.
> 
> Can you factor out those settings to a separate config
'instrumented_libraries',
> and then apply it to the recipe using chromium_apply_config? Another benefit
of
> this arrangement is that it allows us to use instrumented libraries with other
> sanitizers.
> 
> (Note that I don't actually know what chromium_apply_config does, just
guessing
> from the name.)

Fair enough. I added it as as separate config and didn't include it in msan but
only the chromium_msan config.

[earthdok, 2014-09-30 11:57:20]

lgtm

On 2014/09/30 10:56:40, kjellander wrote:
> earthdok: PTAL
> 
>
https://codereview.chromium.org/612753004/diff/20001/scripts/slave/recipe_mod...
> File scripts/slave/recipe_modules/chromium/config.py (right):
> 
>
https://codereview.chromium.org/612753004/diff/20001/scripts/slave/recipe_mod...
> scripts/slave/recipe_modules/chromium/config.py:340:
> gyp_defs['msan_track_origins'] = 0
> On 2014/09/29 13:25:08, earthdok wrote:
> > msan_track_origins=0 should not be a part of the config. This option should
be
> > chosen individually for each bot configuration as it has a significant
impact
> on
> > performance. We already set a default value for it in common.gypi (currently
> 1).
> 
> That makes sense, I left that variable in the existing config and used the
> default for WebRTC in PS#2.
> 
>
https://codereview.chromium.org/612753004/diff/20001/scripts/slave/recipe_mod...
> scripts/slave/recipe_modules/chromium/config.py:341:
> gyp_defs['use_instrumented_libraries'] = 1
> On 2014/09/29 13:25:08, earthdok wrote:
> > I'm also against adding instrumented libraries to the msan config, because
> > certain configurations do not need them. For example, we plan to have an
MSan
> > bot for PDFium, which has no DSO dependencies.
> > 
> > Can you factor out those settings to a separate config
> 'instrumented_libraries',
> > and then apply it to the recipe using chromium_apply_config? Another benefit
> of
> > this arrangement is that it allows us to use instrumented libraries with
other
> > sanitizers.
> > 
> > (Note that I don't actually know what chromium_apply_config does, just
> guessing
> > from the name.)
> 
> Fair enough. I added it as as separate config and didn't include it in msan
but
> only the chromium_msan config.

You probably meant "only the bot recipe".

[kjellander_chromium, 2014-09-30 12:15:26]

The CQ bit was checked by kjellander@chromium.org

[commit-bot: I haz the power, 2014-09-30 12:16:21]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/612753004/40001

[commit-bot: I haz the power, 2014-09-30 12:16:59]

Committed patchset #2 (id:40001) as 292214