Closed Bug 531870 Opened 15 years ago Closed 6 years ago

Crash data not always containing full list of installed extensions

Categories

(Toolkit :: Add-ons Manager, defect)

Product:
Toolkit
Component:
Add-ons Manager
Platform:
x86
Windows XP
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

()

Status:
RESOLVED INACTIVE

People

(Reporter: sicking, Unassigned)

References

Details

(Whiteboard: [crashkill][notacrash]) 

Spinoff from bug 530968 comment 11

The data that we receive from the crash reporter doesn't always contain a proper list of installed extensions.

Bug 530968 contains two crashes, both crashes were produced on an install with identical set of extensions installed, and in both cases the crash was produced using the crashme extension (so after startup). In one case the list of installed extensions shows up fine, in the other the crash data indicates that no extensions were installed:

All extensions: bp-e919ffb7-d698-4f15-ae01-2e59b2091130
No extensions:  bp-7bbd38ef-ef70-4a2c-85d0-02e662091130


It appears this is more than an occasional occurrence. For the [@strchr] crash signature we see heavy correlation data with the idmmzcc.dll dll being loaded:

     95% (42/44) vs.  11% (1313/11946) idmmzcc.dll

However no correlation shows up for the addon with which the dll came. (see bug 530968 comment 1). Possibly this is due to the fact that all or some of the extensions are missing in the crash data.
Are we able to reproduce this ourselves at all?

The list of installed extensions basically comes from the pref extensions.enabledItems. If that were not correct or reset for some reason then we wouldn't be sending the right list so looking at the value of that on a machine that does this might help.
The two crashes mentioned in comment 0 was produced by Henrik Skupin (whimboo), as per bug 530968 comment 11

Don't know if it can be reliably reproduced.
Right now, I wasn't able to find a way to reliable cause this behavior. I have tried with a couple of situations as what I did when I saw the missing extensions the first time but without luck. As you can see from the two mentioned bugs the second one with an uptime of 153s doesn't contain any add-ons. So it's not a start-up crash and the list of extensions should be available.
Are we starting in safe mode in this case? If so, I think we skip annotating the crash report with extension data.
No, Firefox gets started in normal mode after the crash. I don't start the safe mode.
Are any of the crashes that are missing this crashes during startup, before the annotation has been added?
At least not in the case described in comment 0. See comment 0 :)
Whiteboard: [crashkill] → [crashkill][notacrash]
So for what it's worth, when looking at crash data for the crash in bug 530968, it looks like the vast majority of crashes are missing extension completely, but that most of them are startup crashes (uptime in the order 0-10 seconds).

So the incident in comment 0 *might* not be the reason we're not getting addon correlations in bug 530968
If it would help, we could add another annotation, like Startup=1 immediately at app startup, and then set it to Startup=0 from the browser chrome onload, as a rough gauge of "crashed in startup". (If you'd like this, please file a separate bug.)
Depends on: 533234
https://crash-stats.mozilla.com/report/index/bp-f033f0c0-10f4-4c03-8487-a20d72100601 and https://crash-stats.mozilla.com/report/index/9e8f93f1-f2f8-4206-8841-707fd2100625 (not startup crashes) claim no extensions and no processing errors, yet per module list there are clearly extensions such as cal, mintotray, enigmail.  However, most @xtolong crashes list extensions, so these 2 examples are not representative a widespread problem of not listing extensions with this crash sig.
(In reply to comment #4)
> Are we starting in safe mode in this case? If so, I think we skip annotating
the crash report with extension data.

Townsend wrote me:
"I suspect that we put the expected list of extensions in the crash report even in safe mode, though I'd have to check to verify that. It would probably be easy to add whether the app was in safe mode or not to the report though."

Dave, can you check if that's true?  If it is true we may need a new bug - I didn't find one.

Dave also wrote "You can get hints from the module list but that only really counts for extensions that include binary components. One nice side effect of the recent XPCOM registration changes is that its now impossible for an extension to be loaded and doing anything before the extension manager has filled out the extensions list for breakpad so in theory that list that we send to socorro should be correct. Safe mode is I think the exception to this."
Per policy at https://wiki.mozilla.org/Bug_Triage/Projects/Bug_Handling/Bug_Husbandry#Inactive_Bugs. If this bug is not an enhancement request or a bug not present in a supported release of Firefox, then it may be reopened.
Status: NEW → RESOLVED
Closed: 6 years ago
Resolution: --- → INACTIVE
You need to log in before you can comment on or make changes to this bug.