127.57 KB, image/png
94.03 KB, image/png
930.36 KB, image/png
2.43 MB, text/plain
3.86 KB, text/plain
Created attachment 8672638 [details] tmp_7076-Screenshot_2015-10-12-17-42-26-1301930379.png I usually use Nightly, updated every day, to browse the web from my phone (Samsung Galaxy S5, Android 5). It occasionally crashes, but generally it is good enough for browsing, and I hope that tged crash reports are useful (they are, right?) In the last couple of weeks, however, Nightly is barely usable as a day to day browser. On all sites that I read, I see text disappearing, and big black squares appearing. Scrolling up and down sometimes makes the text come back, but after a few moments it disappears again. Often after a minute or so of this happening, Nightly crashes, but I can't be sure that it's related, because it crashes for a lot of reasons ;) In any case, these black squares are much more disruptive than the crashes. I cannot give more precise instructions to reproduce, but I can say that it happens on a lot of different sites: cnn.com, BBC.com, The Atlantic, Quartz, Wikipedia, ynet.co.il, meduza.io, Medium and others.
Thanks for reporting, Amir. Which device are you using?
Samsung Galaxy S5, Android 5.
I am able to reproduce it with Nightly on Galaxy S4, Android 4.2.2. Amir, can you please provide some crash IDs from crashes that happens after this behavior? At least on my case, I suspect it happens when the device is low in memory.
fwiw, I have a GS5, Android 5, and I have not seen this issue.
Some crash reports that could be relevant: ID: 23535e77-0635-423b-8b92-5bec62151015 Signature: OOM | small ID: 861df319-876d-40ec-853a-fce1b2151015 Signature: mozilla::layers::CanvasClientSharedSurface::Updated The second happened shortly after the first, on the same web page.
Snorp, is this your tile size change
It could be, yeah. The fix (from bug 1209801) should be in today's nightly. Amir, does it happen with the latest Nightly?
I just updated, I'll probably know in a day or two.
It still happens, although now the squares are more frequently white than black.
Happened again, this could be related: ID: 0d41e535-32ca-48b9-a6e6-8b8ed2151019 Signature: mozilla::layers::CanvasClientSharedSurface::Updated As I wrote above, now it's more about text disappearing than about black squares covering it, but the general feeling is the same.
And another one: ID: 0d41e535-32ca-48b9-a6e6-8b8ed2151019 Signature: mozilla::layers::CanvasClientSharedSurface::Updated This time it had black squares shortly before it crashed.
Sorry, sent the same ID. Here's the right one: ID: d1bd035b-1519-4767-a34d-2fc842151019 Signature: java.lang.IllegalStateException: Unable to create layer for TabsPanel$TabsLayoutContainer at android.os.MessageQueue.nativePollOnce(Native Method)
(In reply to Michael Comella (:mcomella) from comment #4) > fwiw, I have a GS5, Android 5, and I have not seen this issue. I've started to see this issue on my GS5.
Given the canvas crashes, maybe this is somehow related to bug 1210444 too.
APZ won't affect tiling behaviour and I don't think it will help here. You can also build with --enable-android-apz and try it now, no need to wait for bug 1206872.
Created attachment 8684397 [details] tmp_29816-Screenshot_2015-11-06-22-18-19733637316.png Still happening with the current nightly.
This problem can be reproduced very easily by scrolling up and down on http://www.theverge.com/ You also get the white strip at bottom on that site. FF 45 lollipop 5.0.2 moto g 2nd Gen
This looks very similar to bug 1222810, which as I explained on that ticket could very well be the same OOM issues causing lots of crashes. Amir, could you please attach a logcat captured while reproducing the issue.
Two things: 1. Sorry, I don't know how to do to a logcat. Willing to learn. 2. Actually, I couldn't experience this "black squares" issue recently. It _may_ be fixed somehow, and maybe not.
Actually, it does still happen on today's Nightly.
Another crash of this kind: 1f2552f9-0b3c-434b-b3b8-561a22151124
Yeah, this does look like another instance of OOM.
Created attachment 8695401 [details] white&slowrender.png This is how it appears for me. No black area or crashes. Patches of whites. And text sharpening also lags a lot. Attachment depicts all these issues. All you need to do reproduce this is scroll to the bottom of the page and scroll all the way up. Easily & consistently reproducible on http://www.theverge.com/ Device: Moto G 2nd Gen Lollipop 5.0.2 Firefox Nightly 45(2015-12-03) Happens on Firefox beta 43.0b7 as well
(In reply to bull500 from comment #24) > Created attachment 8695401 [details] > white&slowrender.png > > This is how it appears for me. > No black area or crashes. > Patches of whites. And text sharpening also lags a lot. > Attachment depicts all these issues. > > All you need to do reproduce this is scroll to the bottom of the page and > scroll all the way up. Easily & consistently reproducible on > http://www.theverge.com/ > > Device: Moto G 2nd Gen Lollipop 5.0.2 > Firefox Nightly 45(2015-12-03) > Happens on Firefox beta 43.0b7 as well I've filed bug 1230085 for this.
(In reply to bull500 from comment #24) > Patches of whites. And text sharpening also lags a lot. These effects (the bottom two in your screenshot, at least, maybe the top two as well) are a result of us painting theverge.com very slowly. I filed bug 1230320 for this.
If you want to see this bug live, just go to www.overclockers.ru page and browse some pages (open in new pabs), zoom in-out. FF beta becomes laggy, then after some time it will happen + crash. HTC One M8
Couldn't reproduce with latest FF beta. FF becomes very laggy but I did not saw the white squares and crash...
Actually it crashed when the phone was switched off.
Created attachment 8722748 [details] GS4 crash local build: logcat output I hit this issue. Notably, it crashed immediately after tapping the 3-dot menu in the toolbar.
Created attachment 8722751 [details] GS4 crash local build: abridged logcat output Here's the bit that I think is relevant. Looks like it might be OOM as suspected – here's one of the first notable logs: 02-23 15:51:40.961 E/IMGSRV (11287): :0: PVRPMapKMem: mmap(2) failed: Mapping handle=0x10d: Out of memory
Snorp, I added a logcat from my crash – does this help? Can we make more movement here? Extra detail: I was running Margaret's always reader extension: https://addons.mozilla.org/en-US/android/addon/always-reader/
Looks like another instance of OOM. We're working on that in bug 1247554 and bug 1164027
Nothing really actionable here right now unfortunately.