To hide bug 1565525 from non-Fission users, re-insert a pref check, but for fission.autostart here:
https://hg.mozilla.org/mozilla-central/rev/91d0c9066fd1#l5.55

https://treeherder.mozilla.org/#/jobs?repo=try&revision=3f97a563639a24c17cf087190b0469c9e3e705f0

https://treeherder.mozilla.org/#/jobs?repo=try&revision=2bc672351f5311de0d9cec4ab3195533e845e32f

Sigh. It looks a lot like this patch makes bug 1571254 perma-orange. I wonder if there's a problem with cross-process pref syncing.

Times out even with the compositor in the chrome process...

I don't see the test flipping the pref.

The test passes in the M-fis set.

https://treeherder.mozilla.org/#/jobs?repo=try&revision=643b16985239fa85f1360fe263838379ed2b54ee

Please make sure this change doesn't regress bug 1548687 in the fission-disabled case. I suspect it will since we use those matrices outside of fission.

Or to put it differently, disabling this test if the pref is not set means we lose meaningful test coverage, and I would probably r-minus it (if I were around and had time to fix this properly)

I left a comment on bug 1565525 with a suggestions for other possible fixes that I believe would work. I don't think I'll be able to attempt them myself in the near future though.

(In reply to Kartikaya Gupta (email:kats@mozilla.com) (away until Feb-2020) from comment #9)

Please make sure this change doesn't regress bug 1548687 in the fission-disabled case. I suspect it will since we use those matrices outside of fission.

To clarify a bit more here, the way to ensure this doesn't regress bug 1548687 would be to (a) apply your patch, (b) remove the workaround (this line), and then (c) follow the STR in bug 1548687 comment 0 from step 2 onwards.

(In reply to Kartikaya Gupta (email:kats@mozilla.com) (away until Feb-2020) from comment #9)

we use those matrices outside of fission.

Ah, I didn't realize that.

(In reply to Kartikaya Gupta (email:kats@mozilla.com) (away until Feb-2020) from comment #11)

I left a comment on bug 1565525 with a suggestions for other possible fixes that I believe would work. I don't think I'll be able to attempt them myself in the near future though.

I can give your suggestion a try.

(b) remove the workaround (this line)

Do you know why that workaround is still in-tree?

I requested removal of the workaround here and bug 1565235 was filed for it but I guess nobody has gotten around to writing the patch. I don't think there's anything actually blocking it.

Thanks. With the patches over at bug 1565525, this bug is obsolete.