1399958 - Firefox suddenly stops rendering pages

Status: RESOLVED WORKSFORME

(Core :: Graphics, defect, P1)

Description

[Marco Castelluccio [:marco]]

Attachment: Screenshot

While browsing, after a while, pages start being rendered completely in black.
This is a very recent regression.

I have layers.acceleration.force-enabled set to true.

Comment 1

[Marco Castelluccio [:marco]]

I can reproduce with layers.acceleration.force-enabled set to false too.

Comment 2

[Armen [:armenzg]]

This also affects me on Mac.
I think I noticed yesterday.
I have to copy/paste the URLs into another tab and try again.

Comment 3

[Marco Castelluccio [:marco]]

[Tracking Requested - why for this release]: Very noticeable regression in Firefox 57.

Comment 4

[Julien Cristau [:jcristau]]

tracking as late breaking regression in nightly

Comment 5

[Milan Sreckovic [:milan] (needinfo for best results)]

We'll need a bit more info to act on this.  Right now, it seems OS X and Linux.
For those that see it: what does your about:support graphics section look like when this happens?  Any particular workflows, number of tabs/windows, anything else that correlates?

Comment 6

[Armen [:armenzg]]

Attachment: armenzg about support

ATM I don't have STR. It's happened a handful of times. I will let you know when I discover a pattern.

Comment 7

[Marco Castelluccio [:marco]]

Attachment: marco about:support

(In reply to Armen [:armenzg] from comment #6)
> ATM I don't have STR. It's happened a handful of times. I will let you know
> when I discover a pattern.

Same here.

Comment 8

[Milan Sreckovic [:milan] (needinfo for best results)]

Thanks.  We'll keep tracking this, but at the moment don't have enough information to move this forward.

Comment 9

[Marco Castelluccio [:marco]]

I haven't mentioned it, but I started seeing it yesterday. So the regression must be caused by something that landed very recently.

Comment 10

[Milan Sreckovic [:milan] (needinfo for best results)]

RyanVM's magic gives us this based on comment 9 and the time of filing:
https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=987326974635&tochange=dd6b788f1497

Comment 11

[Milan Sreckovic [:milan] (needinfo for best results)]

if this happens again, and cut-paste-url-to-another-tab workaround works - can you check if the "bad" and the "good" tabs have different content process IDs?

Comment 12

[Milan Sreckovic [:milan] (needinfo for best results)]

Probably not a bad idea for Bill to know about this given bug 1397956 is in the range above.

Comment 13

[Bill McCloskey [inactive unless it's an emergency] (:billm)]

Bug 1397956 puts us in the same situation we were in as recently as August 10 (before bug 1384336 landed). I guess there could be some weird interaction with something that landed recently, but that doesn't seem too likely to me. This could be caused by almost anything.

Comment 14

[Milan Sreckovic [:milan] (needinfo for best results)]

This keeps happening?

Comment 15

[Marco Castelluccio [:marco]]

It hasn't happened again since 2017-09-15.

Comment 16

[Armen [:armenzg]]

No more incidents since the bug was filed.

Comment 17

[Peter Chang[:pchang]]

Based on comment 15 and comment 16, switch to "works for me". Reporter, feel free to open it if you have concern.

Comment 18

[Julien Cristau [:jcristau]]

I'll call this fixed in 57 then.