937957 - OOM Error "Sound Decoding Error" will display sometimes when loading the Songs of Diridum Web Audo Demo from Goo

Status: NEW

(Core :: Web Audio, defect, P3)

Description

[Marc Schifer [mschifer]]

Windows 8.1 - Icona W5 Tablet  ( 1.8Ghz Atom ) 2GB RAM  32Bit.

Error "Sound Decoding Error" will display sometimes when loading the Songs of Diridum  Web Audo Demo from Goo.

If the error dialog is displayed, one or more of the instruments in the instrument control panel will not be displayed when clicking on the panel.

Nightly 28.0a1 (2013-11-12)

- Load the Songs of Diridum Web audio demo
- Watch for the Sound Decoding Error dialog to be displayed.
-- If dialog is not displayed on initial load: Reload page
- Click the instrument control panel
-- Note that for each time the error dialog was displayed, an instrument will be missing from the panel.

- Audio may also have 'crackling' noise during play back

Comment 1

[Maire Reavy [:mreavy]]

I think there are at least 2 bugs here.  One is the decoding error message and another is the crackling (which is likely us running out of CPU and may be a dupe of an existing bug).

Karl, Paul -- I know you are both in NZ this week.  Can you try to repro and give your thoughts on what's going on here?  If the crackling part of this report is a dup, can you reference the bug number?  

If one of you prefers to take ownership of this bug (or on the flip side, if either of you just don't have the time to look at it), let me know.  Otherwise I'll just assign it to one of you -- or perhaps I'll ask JW.  Thanks.

Comment 2

[Paul Adenot (:padenot)]

It would be very useful to know what codec is used, when running the Song of Didirium demo on this tablet. Based on the answer, we can either look into it ourself, or ask someone else to jump in (in the case the this demo is using mp3, and therefore using platform decoders). Also, can this be repro-ed on desktop Windows 8.1?

For the crackling, this is taken care of in other bugs.

Comment 3

[Paul Silaghi, QA [:pauly]]

This is reproducible on x86 Win and Ubuntu desktop platforms.
In browser console:
uncaught exception: out of memory
"No posteffect refs in project" goo.js:1003
An unknown error occured while processing decodeAudioData. index.html
"Decode Error: " [object Object] [object ArrayBuffer] SourceFactory.js:144
An unknown error occured while processing decodeAudioData. index.html
"Decode Error: " [object Object] [object ArrayBuffer] SourceFactory.js:144
An unknown error occured while processing decodeAudioData. index.html
"Decode Error: " [object Object] [object ArrayBuffer] SourceFactory.js:144

Comment 4

[Paul Silaghi, QA [:pauly]]

(In reply to Paul Adenot (:padenot) from comment #2)
> For the crackling, this is taken care of in other bugs.
Can you please post the bug number?

Comment 5

[Maire Reavy [:mreavy]]

Paul S -- Do you still see this problem on Nightly?

Paul A -- Can you post the bug number for the crackling (Comment 4)?

Making P1 since games needs this.

Comment 6

[Maire Reavy [:mreavy]]

(In reply to Paul Adenot (:padenot) from comment #2)
> It would be very useful to know what codec is used, when running the Song of
> Didirium demo on this tablet. Based on the answer, we can either look into
> it ourself, or ask someone else to jump in (in the case the this demo is
> using mp3, and therefore using platform decoders). Also, can this be
> repro-ed on desktop Windows 8.1?
> 

Paul S -- Can you find out what codec is used on the demo?  Once I know if the problems is happening on Nightly (see Comment 5) and what codec is used, I'll find an owner.  And to Paul A's point, it may be someone in media playback (not web audio).  Thanks!

Comment 7

[Paul Silaghi, QA [:pauly]]

(In reply to Paul Adenot (:padenot) from comment #2)
> Also, can this be repro-ed on desktop Windows 8.1?
(In reply to Maire Reavy [:mreavy] from comment #5)
> Paul S -- Do you still see this problem on Nightly?
Repro on 29.0a1 (2013-12-12), Win 7 x86, Win 8.0 x86, Win 8.1 x86.
Even crashed a few times:
bp-9cd01e13-41f9-48fe-a741-05c292131213
bp-4d479c4c-2183-4fa6-90cd-d80582131213
bp-c80b6aad-1031-4a34-abb7-5c22c2131213
bp-b235741d-ab95-4fd1-82e6-5e1912131213

(In reply to Maire Reavy [:mreavy] from comment #6)
> Paul S -- Can you find out what codec is used on the demo?
I'm not sure how to do this. I don't see anything related in the page source or in the console.
If it helps, the demo sound will no longer be heard if setting media.webaudio.enabled to FALSE in about:config.

Comment 8

[Paul Silaghi, QA [:pauly]]

(In reply to Paul Silaghi, QA [:pauly] from comment #7)
> Even crashed a few times:
> bp-9cd01e13-41f9-48fe-a741-05c292131213
> bp-4d479c4c-2183-4fa6-90cd-d80582131213
> bp-c80b6aad-1031-4a34-abb7-5c22c2131213
> bp-b235741d-ab95-4fd1-82e6-5e1912131213
Actually the crashes are handled in bug 943948, I think.

Comment 9

[Paul Adenot (:padenot)]

(In reply to Maire Reavy [:mreavy] from comment #5)
> Paul S -- Do you still see this problem on Nightly?
> 
> Paul A -- Can you post the bug number for the crackling (Comment 4)?
> 
> Making P1 since games needs this.

923319

Comment 10

[Maire Reavy [:mreavy]]

Do we still see the decoding error message?  If so, let's make this bug about that since the other issues are handled by other bugs.

If we're still seeing those messages, I'll assign a bug owner for this.  Thanks.

Comment 11

[Paul Silaghi, QA [:pauly]]

(In reply to Maire Reavy [:mreavy] from comment #10)
> Do we still see the decoding error message?
Yes, confirmed in nightly 29.0a1 (2014-01-29), win 8.1 x86

Comment 12

[Maire Reavy [:mreavy]]

Karl -- Can you looking into what's causing the decoding error message?   Thanks.

Comment 13

[Tony Young [:rfw]]

I've been trying to reproduce this bug using the latest nightly builds (2014-02-14) on both OS X and Windows but haven't managed to get the decoding error after quite a few attempts -- I don't know if this bug still exists.

Comment 14

[Paul Silaghi, QA [:pauly]]

This bug is reproducible only on x86 Windows and Linux platforms.

Comment 15

[Marc Schifer [mschifer]]

I can still reproduce this bug on the hardware I first reported it on 30.0a1

Comment 16

[Karl Tomlinson (:karlt)]

The decode error is most likely due to being out of memory.
The console reports "out of memory" before the decodeAudioData errors:

uncaught exception: out of memory
"No posteffect refs in project" goo.js:1003
An unknown error occured while processing decodeAudioData. mozlod
"Decode Error: " {folder: "music", file: "DoubleBass", gain: 2, options: Object} ArrayBuffer {byteLength: 1242952} SourceFactory.js:144
An unknown error occured while processing decodeAudioData. mozlod
"Decode Error: " {folder: "music", file: "JazzDrums", gain: 2, options: Object} ArrayBuffer {byteLength: 1504830} SourceFactory.js:144
An unknown error occured while processing decodeAudioData. mozlod
"Decode Error: " {folder: "music", file: "Piano", gain: 2, options: Object} ArrayBuffer {byteLength: 1203339} SourceFactory.js:144
An unknown error occured while processing decodeAudioData. mozlod
"Decode Error: " {folder: "music", file: "Saxophone", gain: 2, options: Object} ArrayBuffer {byteLength: 1427676} SourceFactory.js:144

This is consistent with the crashes in comment 7.

I haven't reproduced the decode error by limiting address space on x86_64, but the crashes happen when reloading in a limit of 5GB.

Comment 17

[Karl Tomlinson (:karlt)]

Hitting the 5GB limit in comment 16 was mainly due to 64 MB stack sizes, which would not be appropriate on x86_32 systems.

Unsetting the stacksize limit (for default pthread stack sizes of 2MB) reduces the 32-bit Firefox virtual memory usage to 1.6GB.  Resident memory seems stable at around 920 MB.

Comment 18

[Maire Reavy [:mreavy]]

Mark, Karl - is this bug (the OOM portion) still relevant? (The crackling was split off to another bug.) and if so is it still blocking games?  Thanks

Comment 19

[Karl Tomlinson (:karlt)]

Changes for bug 1199559 and bug 1198656 mean that 43 uses half as much memory setting up the decoded AudioBuffers.

Comment 20

[Karl Tomlinson (:karlt)]

but I wasn't able to verify that resolved the bug, because I didn't reproduce with 41.0.1 either.

Comment 21

[Marc Schifer [mschifer]]

I can still verify this on both 41.0.1 (release) and 44.0a1 20151014030223 (nightly) on the Icona W5 Tablet  ( 1.8Ghz Atom ) 2GB RAM  32Bit. 

I don't think this is a blocker unless we make low memory 32 bit windows tablets a blocking platform.

Comment 22

[Karl Tomlinson (:karlt)]

Fixing bug 1065417 may help here.

FWIW I was testing with a 2GB RAM Pentium U4100 1.3GHz dual core Win7 system.

Comment 23

[Bulk Bug Changes for mreavy's org]

Mass change P2->P3 to align with new Mozilla triage process.