1244504 - Switch trychooser to only use the aliases for all chunks for browser-chrome, devtools-chrome, and their e10s variants

Status: RESOLVED FIXED

(Release Engineering :: General, defect)

Description

[Phil Ringnalda (:philor)]

We do chunk-by-runtime for browser-chrome and devtools-chrome. That means that absolutely every single time that even one test file is added or removed from either suite, every single test might move into a different chunk, as we recalculate what needs to go where to get relatively even runtimes between chunks.

As a result, there is only one single case, where you've pushed a patch already and found where the single test you are interested in ran and then didn't rebase and didn't add another test and didn't remove another test, where choosing to run a particular chunk rather than all chunks is sane behavior. If in an inattentive moment you rebase to the tip of the tree you work off of, boom, the chunk you are running starts passing and you think you fixed your test when really it moved into a chunk you didn't run. If you got backed out for breaking a test in mochitest-e10s-devtools-chrome-5, so you update to the tip of fx-team and reapply your patch and push to try choosing just that one, you probably didn't run that test since someone added three tests, someone else removed two, and because they like to thrash, someone else moved fifty tests into new directories between the push where you got backed out and the one you reapplied your patch on top of.

Comment 1

[Phil Ringnalda (:philor)]

Attachment: Fix

Bonus: this gets rid of the extreme weirdness of having only the alias for non-e10s bc, the alias plus chunks for e10s bc, the alias plus chunks for non-e10s dt, only the alias for e10s dt.

Comment 2

[Phil Ringnalda (:philor)]

Ah, clarity dawns while looking at how wrong the "(foo only)" labels are: that's not an alias for dt, that's the single chunk we started with, then we chunked on Linux but not elsewhere, then we chunked more on debug, etc.

Comment 3

[Joel Maher ( :jmaher ) (UTC -8)]

Comment on attachment 8716782 [details] [diff] [review]
Fix

Review of attachment 8716782 [details] [diff] [review]:
-----------------------------------------------------------------

this is great!  thanks for pushing on this and making the patches.

Comment 4

[Phil Ringnalda (:philor)]

Comment on attachment 8716782 [details] [diff] [review]
Fix

https://hg.mozilla.org/build/tools/rev/7f3bde583d7a