1155138 - Hang while playing 2 flash videos

Status: RESOLVED DUPLICATE of bug 1171948

(Core Graveyard :: Plug-ins, defect)

Description

[Catalin Varga [QA][:VarCat]]

Environment:

FF 37.0.2
Build Id: 20150415140819
Graphics: AMD Radeon HD 5450
OS: Win 8.1 x86

STR:

1. Open 2 youtube videos and set them to run at 1080( make sure flash plugin is used).
2. Switch fullscreen between the 2 videos.

Issue:
After several switches between fullscreens Firefox hangs.

Note:
I've collected the dumps based on the following guide https://wiki.mozilla.org/Flash/Hang_Debugging.This is the dropbox link of the coresponding dumps https://www.dropbox.com/s/7ekkmkxdj38s9gx/dumps.zip?dl=0

Comment 1

[Catalin Varga [QA][:VarCat]]

More accurate STR:

1. Open 2 youtube videos and set them to run at 1080( make sure flash plugin is used).
2. Activate fullscreen on the first video.
3. Exit fullscreen on the first video.
4. Activate fullscreen on the second video.
5. Exit fullscreen on the second video.
6. Repeat steps  2 - 5 untill Firefox stops responding.

Comment 2

[[SV Manager] Florin Mezei, QA (:FlorinMezei)]

[Tracking Requested - why for this release]: Nominating for tracking as it may help identify and fix part of the Flash hangs tracked in bug 789379.

Comment 3

[Ben Turner (not reading bugmail, use the needinfo flag!)]

Attachment: Stacks

Comment 4

[(No longer employed by Mozilla) Aaron Klotz]

I think I've seen this before under different conditions. In both cases the plugin-container was attempting to send NPN_Evaluate to the browser.

Comment 5

[Sylvestre Ledru [:Sylvestre]]

Chris, ni to make sure you see it as you have been looking for str for the flash bugs.


I am tracking it but I am pretty sure that it is going to be a wontfix for 38 :/

Comment 6

[Steven Michaud [:smichaud] (Retired)]

Possibly related to bug 1155976?  See particularly bug 1155976 comment #5.

Comment 7

[Jim Mathies [:jimm]]

(In reply to Steven Michaud from comment #6)
> Possibly related to bug 1155976?  See particularly bug 1155976 comment #5.

no, this is a windows message deadlock. also there isn't any plugin code in the browser stack.

Comment 8

[Chris Peterson [:cpeterson]]

@Catalin: can you reproduce this Flash hang in Firefox Nightly? I believe Aaron landed some Windows message loop fixes in Nightly 40 that have not (yet) been uplifted to Aurora 39.

@Jeromie: has the Flash Player Beta release with Aaron's Windows message loop fixes been released yet? Is this hang reproducible in Flash Player Beta?

Comment 9

[Catalin Varga [QA][:VarCat]]

Hi Chris I've reproduced the issue on Nightly and DevEdition:

FF 40
Build Id: 20150423030204

To be on the safe side that this is the same issue I've collected dumps for the Nightly build.
You can check the related dumps at https://www.dropbox.com/s/radix7qwkbzvzvs/Dumps%20nightly.zip?dl=0

Comment 10

[Sylvestre Ledru [:Sylvestre]]

I am afraid we won't be doing anything in 38. However, tracking in 39.

Comment 11

[Jeromie Clark]

@Chris - AFAIK, all of the fixes on the Flash Player side are in the current betas.  I'm not aware of Mozilla bug numbers for the fixes on the Firefox side, so I don't have any insight there.  I've been operating on the assumption that they're in Nightly and are tracking with Firefox 40 through the stability channels.

Comment 12

[Liz Henry (:lizzard) (relman/hg->git project)]

Aaron is this something you're looking at ?  Or should we wontfix it for 39 at this point? 
If there anything useful in Catalin's information we could follow up on this later.

Comment 13

[(No longer employed by Mozilla) Aaron Klotz]

This is effectively a dupe. I'm going to dupe this one in favor of bug 1171948 because that one has better STR. I think it is important to the 39 release.

Comment 14

[(No longer employed by Mozilla) Aaron Klotz]

I got this to repro. We're deadlocked because the parent process is waiting on the child process to handle some messages. Something isn't working correctly with window neutering because these messages should be getting through.

Comment 15

[Jim Mathies [:jimm]]

(In reply to Aaron Klotz [:aklotz] (please use needinfo) from comment #14)
> I got this to repro. We're deadlocked because the parent process is waiting
> on the child process to handle some messages. Something isn't working
> correctly with window neutering because these messages should be getting
> through.

Is the main browser thread in an ipc out call?

Comment 16

[(No longer employed by Mozilla) Aaron Klotz]

Yes. In this situation that I saw, we had an interrupt race with both sides in outcalls. The plugin-container thread was stuck waiting on a response from the parent and its MessageChannel had deferred the parent's incall. Unfortunately the parent thread was stuck on SetWindowLongPtr(GWL_STYLE) in nsWindow::SetNativeData().