[Bug 24068] New: Crash in dbghelp on 64-bit (WoW)

[wine-bugs at winehq.org]

http://bugs.winehq.org/show_bug.cgi?id=24068

           Summary: Crash in dbghelp on 64-bit (WoW)
           Product: Wine
           Version: 1.3.1
          Platform: x86-64
        OS/Version: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: dbghelp
        AssignedTo: wine-bugs at winehq.org
        ReportedBy: adys.wh at gmail.com


See attached backtrace. Note: I cannot repro the crash that got me there in the
first place, so I won't be able to give further logs. However, I was hinted on
IRC the backtrace could contain enough data to figure the bug out.

[06:34:13] <KittyCat> looks like it's unwinding the stack, so it might just be
a thrown exception. can't tell if it's unhandled or not..
[06:35:33] <@Thunderbird_> dwarf2_virtual_unwind does nothing with context it
just passes it through at the end of the call to this helper function (perhaps
it should have checked context before? I don't know the code)
[06:40:56] <KittyCat> hmm. this doesn't look like it should even compile..
[06:42:08] <KittyCat> oh, wait. okay
[06:43:08] <KittyCat> it's dereferenced so it can be copied to a temp and
restored later
[06:43:42] <KittyCat> and some of the switch cases may do something with it
[06:44:44] <KittyCat> might have to ask Eric Pouech what's supposed to happen
if it's given a NULL context pointer
[06:46:11] <KittyCat> or AJ
[06:55:56] <KittyCat> looks like that code was added in commit
036392c5d131d73e84ad0ed52a796b2cb9c1139d, on March 27th
[07:02:13] <KittyCat> it seems, before that commit it'd have fallen back to a
default_unwind() function, which would also take and use the context (unless
sw_read_mem failed)
[07:06:03] <KittyCat> it looks like it's calling StackWalk explicitly
[07:09:08] <KittyCat> from the looks of it, wow is calling StackWalk with a
NULL context. msdn doesn't seem to have StackWalk, but according to
StackWalk64:
[07:09:14] <KittyCat> "This parameter is required only when the MachineType
parameter is not IMAGE_FILE_MACHINE_I386. However, it is recommended that this
parameter contain a valid context record."
[07:09:30] <KittyCat> and MachineType is being set to IMAGE_FILE_MACHINE_I386
[07:20:46] <KittyCat> from glancing at related commits, it doesn't look like
StackWalk ever properly handled a NULL context parameter
[07:21:46] <KittyCat> though it wasn't until about March 24th that it was
actually used

-- 
Configure bugmail: http://bugs.winehq.org/userprefs.cgi?tab=email
Do not reply to this email, post in Bugzilla using the
above URL to reply.
------- You are receiving this mail because: -------
You are watching all bug changes.