970002 - [OMTC] Browser eats all RAM on http://lab.hakim.se/linjer/#diagonal

Status: RESOLVED WORKSFORME

(Core :: Graphics: Layers, defect)

Description

[Mayank Bansal]

Attachment: memory-report.json.gz

User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:30.0) Gecko/20100101 Firefox/30.0 (Beta/Release)
Build ID: 20140208030207

Steps to reproduce:

1. Create new profile
2. Enable OMTC
3. Go to http://lab.hakim.se/linjer/#diagonal
4. Add some nodes to the demo.
5. Watch memory usage. It continously increases, finally reaching ~2.6GB,  at which point the browser crashes.

I stopeed the demo at ~2.2GB , and opened a about:memory page. I clicked reduce memory 4-5 times. The memory use did not decreae, the browser still used ~2.2 GB.


Actual results:

RAM use continously grows, and eventually the browser will crash


Expected results:

no crash.

FYI, this demo is much smoother in Nightly, than on Chrome34. So great work guys!

Comment 1

[Mayank Bansal]

Attachment: about support.txt

Comment 2

[Mayank Bansal]

crash report:

https://crash-stats.mozilla.com/report/index/be8988ad-b9b1-415e-a9a9-140b02140209

Comment 3

[Alice0775 White]

Regression window with force layers.offmainthreadcomposition.enabled = true

Regression window(m-c)
Good:
http://hg.mozilla.org/mozilla-central/rev/f5b6b5523719
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:27.0) Gecko/20100101 Firefox/27.0 ID:20131003073923
Bad:
http://hg.mozilla.org/mozilla-central/rev/51b36c5fd45f
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:27.0) Gecko/20100101 Firefox/27.0 ID:20131003080100
Pushlog:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=f5b6b5523719&tochange=51b36c5fd45f


Regression window(m-i)
Good:
http://hg.mozilla.org/integration/mozilla-inbound/rev/3cf52700118c
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:27.0) Gecko/20100101 Firefox/27.0 ID:20131002135314
Bad:
http://hg.mozilla.org/integration/mozilla-inbound/rev/477781ec7d38
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:27.0) Gecko/20100101 Firefox/27.0 ID:20131002135412
Pushlog:
http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=3cf52700118c&tochange=477781ec7d38
Regressed by:
477781ec7d38	Nicolas Silva — Bug 916118 - Simplify the TextureClient deallocation protocol. r=nrc

Comment 4

[Bas Schouten (:bas.schouten)]

Nicolas, please have a look at this.

Comment 5

[Nicolas Silva [:nical]]

(In reply to Bas Schouten (:bas.schouten) from comment #4)
> Nicolas, please have a look at this.

yup.
Two gigs of MemoryTextureClient, sounds pretty bad!

Comment 6

[Nicolas Silva [:nical]]

We don't ship on windows with OMTC enabled yet so the different releases are not "affected".

Comment 7

[Mayank Bansal]

FWIW, this doesnt reproduce for me anymore on the latest Nightly

Comment 8

[Nicolas Silva [:nical]]

(In reply to mayankleoboy1 from comment #7)
> FWIW, this doesnt reproduce for me anymore on the latest Nightly

Ok great, We have been making a lot of changes around windows OMTC lately so this probably got fixed in the mix. I'll close this bug as I haven't been able to reproduce either. Also, the canvas-related OMTC code paths will certainly go through a lot more changes before we can ship OMTC on windows.

Feel free to reopen if this bug shows up again.