992324 - disable interruptible reflow in reftest

Status: RESOLVED FIXED

(Testing :: Reftest, defect)

Description

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

I was discussing the corrupted-image reftest failures like bug 930894 with seth, and one theory we came up with was that reflow might be being interrupted.

This would only happen if:
 * something thinks there are pending user events, which is plausible, though probably shouldn't be happening

 * clock skew is causing the interrupt timer to decide to interrupt really quickly, which is also plausible on a VM (and those failures are only happening on the B2G desktop VMs)

Given the lack of pending user events, reftest normally wouldn't be testing interruptible reflow.  So we should probably just explicitly disable interruptible reflow in the reftest harness.  (If we wanted to leave it, we would want to make the state machine in MakeProgress wait for all interruptible reflows to complete, but given that we normally wouldn't expect this to ever happen, it seems better to just disable it.)

Comment 1

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Unless I'm missing something, we don't have any prefs to disable interruptible reflow, only env vars.  Is that right?  And are you ok with me adding a pref?

Comment 2

[Boris Zbarsky [:bzbarsky]]

You're correct, and adding a pref sounds just fine.  I can't recall now why we did this with env vars instead of prefs to start with.

Comment 3

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: patch 1 - Add a pref to disable interruptible reflow

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: patch 2 - Disable interruptible reflow in reftest harness

Without this patch, mInterruptsEnabled is frequently (although not
always) initialized to true in nsPresContext::ReflowStarted while
running reftests; with the patch it is always initialized to false.

Comment 5

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 8404327 [details] [diff] [review]
patch 1 - Add a pref to disable interruptible reflow

r=me

Comment 6

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 8404328 [details] [diff] [review]
patch 2 - Disable interruptible reflow in reftest harness

r=me

Comment 7

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/3645bb7a4a7f
https://hg.mozilla.org/integration/mozilla-inbound/rev/e74883f49f5d

(and try run at https://tbpl.mozilla.org/?tree=Try&rev=8b414d954c01 )

Comment 8

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/3645bb7a4a7f
https://hg.mozilla.org/mozilla-central/rev/e74883f49f5d