1474952 - Symbols in crash stack are sometimes not getting demangled

Status: RESOLVED DUPLICATE of bug 1474871

(Socorro :: General, task)

Description

[Kartikaya Gupta (email:[e-mail])]

e.g. in https://crash-stats.mozilla.com/report/index/e2de2d0a-8734-45a0-8fd0-ff4530180702 the stack has mangled symbol names. There are other crashes from around the same time for which the symbol names are fine, so I'm not sure why some reports have this problem.

Comment 1

[Will Kahn-Greene [:willkg] ET needinfo? me]

I think the next step is for someone to pull the symbols for that crash and see if something happened when generating them or if they don't exist or what.

Here are some other crash reports and build ids that exhibit the same mangling:

https://crash-stats.mozilla.com/report/index/6e8a7fe4-e670-4cd4-881b-b33560180728  20180708100828
https://crash-stats.mozilla.com/report/index/6f1c3a5d-4af1-4836-8559-d176e0180716  20180711221812
https://crash-stats.mozilla.com/report/index/b101b952-7f9b-453b-a91b-4126a0180715  20180711100118

The build ids are all valid build ids, so we should have symbols for them.

Comment 2

[Kartikaya Gupta (email:[e-mail])]

I just realized, these mangled names might not be valid. The symbol in stack frame 2 of the crash report in comment 0 is "_ZN11panic_abort18__rust_start_panic5abort17h8d2f6471e2847f91E". Running that through c++filt just spits out the same thing. If so then maybe rustc is generating badly mangled names, or maybe mangled names that are incompatible with whatever demangler we're using?

Comment 3

[Will Kahn-Greene [:willkg] ET needinfo? me]

I remembered there was a rust demangling bug in the crash reporter recently:

https://bugzilla.mozilla.org/show_bug.cgi?id=1474871

Socorro doesn't do any demangling, so if it's an issue with demangling, that's a problem elsewhere.

Is this effectively a dupe of that bug? You're mentioned in the description:

https://bugzilla.mozilla.org/show_bug.cgi?id=1474871#c0

Comment 4

[Kartikaya Gupta (email:[e-mail])]

That sounds plausible. I hadn't seen that bug before. I'll dupe this over for now and file a new bug in Toolkit/Crash Reporting if I still see this as an issue on new crash reports.