The "green-ness" of the step is not something LogDog will influence. It is based on annotation STEP_FAILED / STEP_WARNING, so I don't think this has anything to do with LogDog.

Note the process returns code 255, which is properly propagated by LogDog and returned as 255, causing the overall build to fail as expected.

generate_build_files didn't fail. The recipe itself failed, it tried to add a config again during the "without patch" steps that was already added during the "with patch" steps. Adding the same config twice is erroneous.

+cc dpranke who added this logic I believe.

When an exception gets raised in the recipe itself it emits this uncaught exception step. Not sure why it's green, that seems like another issue.

Yup, that looks broken, but I don't know that it's related to logdog or perhaps something else.

I agree w/ smut@'s analysis that it looks like the 'ninja' config is getting applied twice; however, either that didn't used to produce an error, or we haven't noticed this for months, which seems unlikely?

I can look into fixing the recipe, but I won't be treating this urgently unless we're seeing this a lot more often now. Maybe someone can look and see if this is a regression elsewhere in the engine as well?

The reason I filed this ticket is because it is an invisible issue :-) It happens pretty often and leads to invalid flakiness reports. This one to be precise: http://crbug.com/594612 

Other examples can be found here: https://chromium-try-flakes.appspot.com/all_flake_occurrences?key=ahVzfmNocm9taXVtLXRyeS1mbGFrZXNyEgsSBUZsYWtlIgdjb21waWxlDA

@pgervais: Yup, agreed. The fact that the recipe is producing weird results is, in some sense, a separate issue that should be fixed (or at least understood) regardless.

Looking...

This is indeed caused by the recipe applying the config twice. It's doing this because it compiles with, and without the patch.

This looks like it is actively breaking the without compile patches:

https://build.chromium.org/p/tryserver.chromium.mac/builders/ios_dbg_simulator_ninja/builds/159953
https://build.chromium.org/p/tryserver.chromium.mac/builders/ios_dbg_simulator_ninja/builds/159952
https://build.chromium.org/p/tryserver.chromium.mac/builders/ios_dbg_simulator_ninja/builds/159942

all have the same exception.

You could just make the ios build function take a 'without_patch' argument that changes the suffix and doesn't apply these configs a second time.

In comment #8, martiniss@ wrote:
> You could just make the ios build function take a 'without_patch' argument that changes the suffix and doesn't apply these configs a second time.

True. There's no doubt that the recipe code is wrong. 

However, separately, this code has been this way for a long time. Any idea why we'd just be noticing this now? Did something else change in the recipe engine that would cause this to start throwing, or something?

Also, do we have any idea why the 'uncaught exception' step is green? That seems a bit weird ...

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/tools/build.git/+/d1e8fec122f867fc1d02e4d38cce2da6c7d44ead

commit d1e8fec122f867fc1d02e4d38cce2da6c7d44ead
Author: dpranke@chromium.org <dpranke@chromium.org>
Date: Fri Apr 15 01:00:24 2016

Fix uncaught exception on ios_dbg_simulator_ninja bots.

R=martiniss@chromium.org, smut@chromium.org
BUG= 603641 

Review URL: https://codereview.chromium.org/1886813005

git-svn-id: svn://svn.chromium.org/chrome/trunk/tools/build@299945 0039d316-1c4b-4281-b951-d872f2087c98

[modify] https://crrev.com/d1e8fec122f867fc1d02e4d38cce2da6c7d44ead/scripts/slave/recipe_modules/ios/api.py

Okay, hopefully fixed now. Please reopen if you see an issue again.