Closed Bug 1245200 Opened 9 years ago Closed 8 years ago

Plugin crash spike in 45.0b1

Categories

(Core Graveyard :: Plug-ins, defect)

Product:
Core Graveyard
Component:
Plug-ins
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

(firefox44 unaffected, firefox45- disabled, firefox46 affected, firefox47 affected, firefox49 affected, firefox50 affected, firefox51 affected)

Status:
RESOLVED WORKSFORME
Tracking Flags:
Tracking Status
firefox44 --- unaffected
firefox45 - disabled
firefox46 --- affected
firefox47 --- affected
firefox49 --- affected
firefox50 --- affected
firefox51 --- affected

People

(Reporter: kairo, Unassigned)

References

Details

Crash Data

Plugin crashes on beta are 3-4x as high after shipping 45.0b1 as they were before (44.0 betas).

Only <15% of the plugin issues have e10s enabled, so the experiment is probably not the cause for this.

Flash 21 beta also just shipped, so in theory it could play a role in the spike as well, but looking at the signatures I so far don't think that's the case, and it's probably changes in 45 vs. 44 that are causing this.
[Tracking Requested - why for this release]:
status-firefox44: --- → unaffected
status-firefox45: --- → affected
tracking-firefox45: --- → ?
Aaron, I know you did some work in the plugin area, do you know if we landed anything in 45 that may be the cause of or related to this issue?
Flags: needinfo?(aklotz)
Telemetry data confirms this jump: https://telemetry.mozilla.org/new-pipeline/crash-summary.html has us going from 0.88 to 1.66 plugin crashes per thousand hours of usage.

Kairo, can you do a straight topcrash comparison between the last 44 beta versus the 45 beta to see if any particular signatures jump out?
Flags: needinfo?(kairo)
Is this consistent across both 32 and 64-bit builds?
Flags: needinfo?(aklotz)
(In reply to Bill McCloskey (:billm) from comment #5)
> I asked about this crash and Benjamin pointed me to this bug:
> https://crash-stats.mozilla.com/signature/?product=Firefox&version=45.
> 0b&date=%3C2016-02-02T22%3A39%3A34&date=%3E%3D2016-01-
> 26T22%3A39%3A34&signature=F_1799236063_______________________________________
> ___________&_columns=date&_columns=product&_columns=version&_columns=build_id
> &_columns=platform&_columns=reason&_columns=address&page=1#reports
Flags: needinfo?(jeclark)
I've opened ADOBE 4113091 on this.  We'll start looking at it ASAP.
Flags: needinfo?(jeclark)
(In reply to Benjamin Smedberg  [:bsmedberg] from comment #3)
> Kairo, can you do a straight topcrash comparison between the last 44 beta
> versus the 45 beta to see if any particular signatures jump out?

The signature that meanwhile has been assigned to this bug is indeed what stands out.

This signature splits between Flash versions as follows:
1 	21.0.0.130 	5477 	56.20 %
2 	20.0.0.294 	2626 	26.95 %
3 	(none)  	1546 	15.86 %
4 	20.0.0.279 	70 	0.72 %
5 	20.0.0.274 	26 	0.27 %

The (none) which actually is empty is an artifact of bug 1245165.
Flags: needinfo?(kairo)
Initial analysis indicates that this is related to the Async Drawing feature that we've been collaborating on.  Matt is investigating.
cc: dvander so that he knows what's up
dvander, what do we know about what's going on here? We need to figure this out or disable this until we figure out what's causing these crashes.
Oh, I see this was already disabled in bug 1246311
FYI, this issue should be resolved in Flash Player 21.0.0.136 or higher.
(In reply to Jeromie Clark from comment #13)
> FYI, this issue should be resolved in Flash Player 21.0.0.136 or higher.

Looking at the "Flash version" section in https://crash-stats.mozilla.com/signature/?signature=F_1799236063__________________________________________________ it at least looks like 21.0.0.136 is still affected.
Bug 1246311  doesn't fix the problem with the latest Flash beta crashing under Windows 10. The latest Flash beta which is 21.0.0.136 works fine until you try to go full screen at some sites like cnn.com.  It doesn't matter if 'dom.ipc.plugins.asyncInit.enabled' is true or false.
Thanks.  We caught it this morning.  We addressed the original crash, but exposed another.  We're looking into it.
Is Adobe Flash Player 21.0.0.146 Beta supposed to fix this problem because it doesn't.
No.  We just learned that the original fix didn't work yesterday.  We're not quite that fast.  :)
Tracking to make sure it is fixed before release day.
tracking-firefox45: ? → +
FYI... Issue not fixed in latest Flash beta v21.0.0.157
If I turn off HWA either in Fx or in Flash, Fullscreen works fine.
To add, switching from Fullscreen to normal screen results in Nightly hanging.  Had to kill the Nightly task.
In crash stats, the issue seems to be gone after 45.0 beta 3
status-firefox45: affected → ---
(In reply to Sylvestre Ledru [:sylvestre] from comment #23)
> In crash stats, the issue seems to be gone after 45.0 beta 3

Yes, due to the disabling of Async Drawing in bug 1246311. As soon as Async Drawing is on again, we'll see those crashes again.
Crash volume for signature 'F_1799236063__________________________________________________':
 - nightly (version 52): 0 crashes from 2016-09-19.
 - aurora  (version 51): 2 crashes from 2016-09-19.
 - beta    (version 50): 454 crashes from 2016-09-20.
 - release (version 49): 1 crash from 2016-09-05.
 - esr     (version 45): 0 crashes from 2016-07-25.

Crash volume on the last weeks (Week N is from 10-17 to 10-23):
            W. N-1  W. N-2  W. N-3  W. N-4
 - nightly       0       0       0       0
 - aurora        2       0       0       0
 - beta        108     167     167       0
 - release       0       0       0       0
 - esr           0       0       0       0

Affected platform: Windows

Crash rank on the last 7 days:
             Browser   Content Plugin
 - nightly
 - aurora                      #87
 - beta                        #5
 - release                     #109
 - esr
status-firefox49: --- → affected
status-firefox50: --- → affected
status-firefox51: --- → affected
As filed this bug isn't helpful. We'll revisit when we get async drawing back.
Blocks: 1229961
Status: NEW → RESOLVED
Closed: 8 years ago
Resolution: --- → WORKSFORME
Product: Core → Core Graveyard
You need to log in before you can comment on or make changes to this bug.