Crash stacks should provide additional information for inlined function calls in stack frames
Categories
(Socorro :: Webapp, task)
Tracking
(Not tracked)
People
(Reporter: kats, Unassigned)
Details
See bug 1280428 comment 9. This is a case where a function got inlined, and the crash stack showed a line number from the inlined function instead of the (much more useful) function that it got inlined into. Markus was able to manually extract this information, and it looks like a set of steps that could be automated without too much trouble. It would be great if crash-stats did this automatically, detecting stack frames that are omitted due to inlining, and inserting them into the stack.
|
||
Comment 1•3 years ago
|
||
My guess is that this is something we'd need to solve in minidump-stackwalk which is doing the symbol lookup.
But I don't think there's enough information in this bug to do anything with. If it's still a problem, it'd help to have a recent crash report along with the details of the inlined function and how you figured that out.
|
||
Comment 2•3 years ago
|
||
I can cite at least two more examples where I had to do this: bug 1599861 comment 3 and bug 1602813 comment 4. There was another one more recently but I didn't save the bug number.
However, this is probably a duplicate of bug 1398533.
|
|
||
Comment 3•3 years ago
|
||
The two bugs are in different parts of our crashreporting system. Bug #1398533 covers fixing the client side of things and building symbols so we have the information to derive inlining from.
This one covers extracting additional information so that on Crash Stats we can show inlining information regardless of whether bug #1398533 gets fixed.
I don't think Socorro can grow this muscle and instead we should fix the rest of the crash reporting machinery to support inlined functions.
Description
•