400714 - Flash crashes on Mac skyrocketed in Firefox 2.0.0.8

Status: RESOLVED INVALID

(Core Graveyard :: Plug-ins, defect)

Description

[Daniel Veditz [:dveditz]]

Mac Flash crashes have skyrocketed in Firefox 2.0.0.8 compared to Firefox 2.0.0.7. Looking at talkback data, even though total crashes in a 10 day window
favor the old version 45K to 27K (we shipped less than 10 days ago, some people take a week to upgrade), crashes in Flash on a Mac have gone from 223
to 844.

FF 2.0.0.7             FF 2.0.0.8
223 Flash crashes      844 Flash crashes
13.5% of Mac crashes   60% of Mac crashes
0.5% of all crashes    3% of all crashes

There were two check-ins to mozilla/modules/plugin/base/, bug 392505 and bug 354595 (the latter suspiciously Mac-only). Of course it might not be a direct plugin change, there were other layout/content changes that might have affected events, timing, or some other triggering condition.

Comment 1

[Mike Schroepfer]

+ for 1.8.1.9 to make sure it stays on the list

Comment 2

[Johnny Stenback (:jst)]

While I can't say for certain that bug 392505 and bug 354595 didn't cause new crashes in plugin code, it is *highly* unlikely that they'd be the cause of any new crashes...

Comment 3

[Daniel Veditz [:dveditz]]

I agree it's unlikely, but they were the handiest straws to grasp :-(

Comment 4

[Daniel Veditz [:dveditz]]

On 10/23/2007 between 0800 and 1700 there were 82 "Flash Player" crashes on mac; 34 were using FF2.0.0.8 (41%). That's a pretty low percentage, people who don't upgrade Firefox might very well be using an old Flash?

On 10/20/2007 in the same hours there were 750 crashes! 52 of the first 100 were using FF2.0.0.8 so not a whole lot more using the latest. Clearly this spike was not caused by a regression in FF 2.0.0.8 since it came and went and affected old versions just as much.

I tried getting numbers on a "before ship" date, but could never get number for 10/10 or 10/16, the talkback site just never returned anything (trying both before and after I got results for the more recent dates). The topcrash result page that was stuck on 10/20 and was supposed to be regenerated is now empty.

Comment 5

[Samuel Sidler (old account; do not CC)]

I did some specific numbers between Firefox 2.0.0.8 and Firefox 2.0.0.7 on equivalent days, there was a huge increase in crashes on Saturday (10/20):

                 10/19	10/20	10/21	10/22	10/23
Firefox 2.0.0.8  113	722	79	86	82
Firefox 2.0.0.7  20	136	16	14	14

There's a clear spike there and I'm not entirely sure why. Speculations are on a bad ad or flash movie that went around on sites. I haven't gone back further, but I can.

Comment 6

[Daniel Veditz [:dveditz]]

This appears to be a spike due to web content since it equally affected older versions of Firefox on closer inspection of the talkback data. The spike has subsided since Saturday. Not blocking the 1.8.1.9 regression-firedrill.

Possibly "WFM" or "INVALID (plugin vendor issue)" but I'll leave that for a later look.

Comment 7

[Daniel Veditz [:dveditz]]

Finally got data pre-2.0.0.8,

8am-5pmPST   Crashes    % using latest
10/16          55         16%
10/20         750         52%
10/23          82         41%

Comment 8

[Samuel Sidler (old account; do not CC)]

This has dropped a lot in 2.0.0.11 and can probably be closed, with the blocking request removed.

Comment 9

[Daniel Veditz [:dveditz]]

We've dropped back down to 40-50 crashes a day, about what it was before the spike and thus not something we regressed with our release. Still a Flash bug.

Comment 10

[Samuel Sidler (old account; do not CC)]

Clearing blocking request.