637858 - Enable (some) reftests for out-of-process content on Fedora machines

Status: RESOLVED FIXED

(Release Engineering :: General, defect)

Description

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Spun off from bug 623613.

To summarize, we want to run these tests with the pref

  browser.tabs.remote=true

and

  MOZ_LAYERS_FORCE_SHMEM_SURFACES=1

in the environment.

We'll get a large benefit from just enabling reftests for even one directory that currently passes all tests, so that we can tell when m-c pushes completely break out-of-process content.

Comment 1

[Armen [:armenzg]]

Do you know the subset or not yet?

Comment 2

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

First candidate would be layout/reftests/reftest-sanity/reftest.list, but I need to check if they're still passing.

Comment 3

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

In the opt run in bug 623613, all the reftest-sanity tests passed.

Comment 4

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

In my last try push (263d877928ac)
 - all the linux reftest-ipcs passed layout/reftests/reftest-sanity/reftest.list
 - both the opt and debug builds failed another 30 or so tests
 - the debug builds also exhibited the leak that's showing up in debug crashtest-ipc

So, I think we're ready to start running just layout/reftests/reftest-sanity/reftest.list on opt builds!  Let me hunt down the runreftest.py incantation for that.

Comment 5

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Looks like in the config we can just specify

+        elif suite in ('reftest-ipc'):
+            return ['--setpref=browser.tabs.remote=true',
+                    'reftest/tests/layout/reftests/reftest-sanity/reftest.list']

Instead of reftest/tests/layout/reftests/reftest.list.

Comment 6

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

The reftest-sanity tests are ready to be enabled.  These will catch wholesale breakage of shadow-layers code, like what happened with the initial landing of bug 626602.

Comment 7

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

found in triage: can you clarify:

1) do you want us to run these tests with changed pref/env settings *and* also run the same test suites without changing pref/env settings? 

2) This bug only mentions Fedora machines. how do you want us to handle testing on win32, osx, android, maemo?

Comment 8

[Armen [:armenzg]]

This got fixed in bug 623613.

With:
https://bugzilla.mozilla.org/show_bug.cgi?id=623613#attachflag6
on March 2nd.