Crash in [@ xul.dll] editing Filters
Categories
(Thunderbird :: Filters, defect)
Tracking
(Not tracked)
People
(Reporter: wsmwk, Unassigned)
References
(Blocks 1 open bug)
Details
(Keywords: crash, Whiteboard: [filtermgmt])
Crash Data
Many users cite editing filters as the cause of their crash. All signatures are OOM or xul.dll. As a result, none have good stacks. And I have not found other bug reports describing this.
Unfortunately not much data.
Crash report: https://crash-stats.mozilla.org/report/index/b0fb7309-71ec-447d-9496-ad6c40230317
MOZ_CRASH Reason: called `Result::unwrap()` on an `Err` value: ThreadPoolBuildError { kind: IOError(Os { code: 8, kind: OutOfMemory, message: "Not enough memory resources are available to process this command." }) }
Top 10 frames of crashing thread:
0 xul.dll xul.dll@0x4e165d1
1 xul.dll xul.dll@0x40de3c6
2 xul.dll xul.dll@0x40de31a
3 xul.dll xul.dll@0x4306ed0
4 xul.dll xul.dll@0x4306b32
5 xul.dll xul.dll@0x4306a88
6 xul.dll xul.dll@0xcef2ae
7 xul.dll xul.dll@0xcefa9e
8 xul.dll xul.dll@0xc213b8
9 xul.dll xul.dll@0xc9327c
Bp-36dbcdb4-e545-4659-ad4c-097e90230317 is from the same person.
Looked at a dozen crashes. All have 4gb virtual memory or less. But not all are OOM. For example these all have plenty of physical and virtual memory:
- bp-12413f8c-538c-47a2-bd1e-0ed530230524 xul.dll | _InternalCallWinProc
- bp-a5919697-ca4d-4b89-a30e-646860230503 xul.dll | _InternalCallWinProc
- bp-69ad09a1-a6a6-45ad-8b7f-82c290230526 thunderbird@0x5a3c2
- bp-5ba9de84-85a3-4ede-aa29-7ef980230526. xul.dll | _InternalCallWinProc
| Reporter | ||
Updated•3 years ago
|
| Reporter | ||
Comment 1•2 years ago
|
||
Thunderbird (And firefox) crash rate dropped dramatically ~October 25, around the time of Bug 1860841 - socorro deploy: 2023.10.24, so stacks may have changed.
For Thunderbird some of the crash signatures may have become Bug 1840573. But not enough information to say more than that.
Comment 2•2 years ago
|
||
Previously, the signature would be xul.dll for crash reports on Windows in the case where the crash reporter couldn't figure out the debug id to put in the minidump. When the Socorro stackwalker went to unwind and symbolicate the stack in the minidump, it would see the xul.dll module, but no debug id, so it couldn't acquire the symbols file in order to symbolicate the frame.
Socorro deploy 2023.10.24 pushed out the the final change to the stackwalker where if it finds itself in the case where it is trying to download the symbols file for a module and it has a code id (a Windows build identifier), but no debug id, it would use a new API enhancement to the Mozilla Symbols Server I wrote that allows it to use the code id to look up the debug id and then retrieve the symbols file. These changes allowed us to symbolicate many crash reports we weren't able to symbolicate before which dramatically improved signatures.
I wrote about it in this blog post: https://bluesock.org/~willkg/blog/mozilla/socorro_tecken_code_info_retro.html
Description
•