1194921 - Cleanup config.py now that esr31 is EOL

Status: RESOLVED FIXED

(Release Engineering :: General, defect)

Description

[Ryan VanderMeulen [:RyanVM]]

Now that esr31 is dead, we can kill a lot of dead code in config.py. While I was at it, I also reorganized things and gave some of the suites better names.

 mozilla-tests/config.py | 1030 +++++++++++++++++++-----------------------------
 mozilla/config.py       |  239 -----------
 2 files changed, 412 insertions(+), 857 deletions(-)

Comment 1

[Ryan VanderMeulen [:RyanVM]]

Attachment: patch

Comment 2

[Ryan VanderMeulen [:RyanVM]]

Comment on attachment 8648331 [details] [diff] [review]
patch

Ben, one thing I specifically wanted to bring to your attention is the removal of the Gecko 34+ OSX signing block now that b2g37 is the oldest branch still building OSX Firefox builds.

Comment 3

[Ryan VanderMeulen [:RyanVM]]

Attachment: builder diff

I know this looks scary, so I'll break down the notable changes:
* Turns out Holly had a bunch of test suites enabled that weren't ever intended to be (due to the ordering of all the random blocks). I confirmed with billm that returning to just mochitests + reftests is fine.
* Reftest-omtc removed. Verified with Milan that it's not needed (never really got used in the first place).
* Split mochitest-dt into 2 chunks by default. This is long-overdue (and we routinely hit 120min timeouts on Linux these days).
* Split reftest-noaccel into 2 chunks on Linux. Same story as the last one.
* Enabled debug Cpptests on Win32. AFAICT, these should have been enabled all along and never were. Chris Manchester confirms that they should pass (and they already are on Win64).
* Re-enable crashtest-e10s on all Linux platforms. Looks like it was accidentally disabled for the most part when reftest-e10s was split into 2 chunks. Oops.

Comment 4

[Ryan VanderMeulen [:RyanVM]]

Hopefully this all makes sense. I think the results config.py files are easier to follow and I hope you agree :-)

Comment 5

[[e-mail] (:bhearsum)]

Comment on attachment 8648331 [details] [diff] [review]
patch

Review of attachment 8648331 [details] [diff] [review]:
-----------------------------------------------------------------

The signing block is definitely right, we killed esr31 branches last week! The rest of mozilla/config.py looks good too - I ran it through dump_master.py, and after factoring out the removal of LD_LIBRARY_PATH, the patch to this file is a no-op.

Because you're doing some chunking at the same time as dead code removal, it's a bit harder to review the tests changes. dump_master is too noisy, so I ran it through builder_list.py instead. There's a lot of differences, but I think they're all expected. Here's the full list if you want to double check: https://gist.github.com/bhearsum/f32854fefc00bcbeb918

Comment 6

[Ryan VanderMeulen [:RyanVM]]

Comment on attachment 8648331 [details] [diff] [review]
patch

Yeah, my apologies for not doing those in a separate patch. Got a bit carried away with things :-)

Thanks for the quick review!

https://hg.mozilla.org/build/buildbot-configs/rev/68a8f95330e4

Comment 7

[Kim Moir [:kmoir] ET]

In production