859527 - Stacks truncated again (using minidump_stackwalk on a Mac Tinderbox debug build)

Status: RESOLVED DUPLICATE of bug 884300

(Toolkit :: Crash Reporting, defect)

Description

[Jesse Ruderman]

The stack in bug 859526 is truncated.  It's relying entirely on stack-scanning and not getting very far (possibly a similar problem to bug 787302, which was fixed).

I'm not sure if this is happening to all crashes, but it seems to affect many.

Comment 1

[Jesse Ruderman]

This limits my ability to ignore crashes selectively.

Comment 2

[Jesse Ruderman]

For example, I can't distinguish between "###!!! ABORT: You can't dereference a NULL nsCOMPtr with operator->()" crashes that occur in different functions, because the stack is wrong past NS_DebugBreak.  With the crash in bug 866839:

 0  libmozalloc.dylib!mozalloc_abort(char const*) [mozalloc_abort.cpp : 30 + 0x0]
    rbx = 0x00007fff7f329c68   r12 = 0x00000001036a5340
    r13 = 0x00000001036d9060   r14 = 0x00007fff5fbfcec0
    r15 = 0x00007fff7f329c68   rip = 0x00000001000aaaa4
    rsp = 0x00007fff5fbfce70   rbp = 0x00007fff5fbfce80
    Found by: given as instruction pointer in context
 1  XUL!NS_DebugBreak [nsDebugImpl.cpp : 387 + 0x7]
    rip = 0x0000000102af7182   rsp = 0x00007fff5fbfcea0
    Found by: stack scanning
 2  XUL!RDFServiceImpl::GetAnonymousResource(nsIRDFResource**)::gChars + 0x6cb
    rip = 0x0000000103806ecc   rsp = 0x00007fff5fbfcea8
    Found by: stack scanning

Comment 3

[(not currently active) Ted Mielczarek]

I might have fixed this in bug 884300, can you try with a new build and see if you get better results?

Comment 4

[Jesse Ruderman]

Yes, I'm getting good stacks now.  Thanks!