Closed
Bug 1307192
Opened 8 years ago
Closed 3 years ago
Getting symbols back from symbolication that don't seem plausible
Categories
(Core :: Gecko Profiler, defect)
Core
Gecko Profiler
Tracking
()
RESOLVED
WORKSFORME
People
(Reporter: fitzgen, Assigned: jimb)
References
(Blocks 1 open bug)
Details
When sfink, jimb, and I sat down and looked at some profiles together, we saw a lot of time in `std::__adjust_heap` or something like that called directly from `__poll` and a bunch of things that we don't even call in gecko. Pretty sure this is a symbolication bug.
Comment 1•8 years ago
|
||
I believe that is from https://cleopatra.io/#report=c1afe580c1be80ec7f4e247a2db8d422d002d8c8&invertCallback=true (select the content process)
Assignee | ||
Updated•8 years ago
|
Assignee: nobody → jimb
Assignee | ||
Comment 2•8 years ago
|
||
Right, in the linked profile, select "Content" from the "Frames" drop-down, and then check "Invert callstack" in the "View Options" area. Open up the "total" item, and "std::__adjust_heap" is the second child, credited with 2.4% of samples. Its callers seem to all be JavaScript code, which is ridiculous.
I assume that's an error introduced at symbolication time. On my own machine I'm not able to get "std::__adjust_heap", but I can reproduce a similar callee which is labeled "Unknown", and also has exclusively JS callers, and is responsible for 4.2% of the samples.
Since it's called from JS, I wonder if it's a trampoline of some sort. Those might not have debugging information.
Updated•3 years ago
|
Status: NEW → RESOLVED
Closed: 3 years ago
Resolution: --- → WORKSFORME
You need to log in
before you can comment on or make changes to this bug.
Description
•