Bug 552619

Summary: gdb crash when xemacs is compiled with -feliminate-dwarf2-dups
Product: [Fedora] Fedora Reporter: Jerry James <loganjerry>
Component: gdbAssignee: Tom Tromey <tromey>
Status: CLOSED ERRATA QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: medium Docs Contact:
Priority: low    
Version: 12CC: jan.kratochvil, patrickm, pmuldoon, tromey
Target Milestone: ---   
Target Release: ---   
Hardware: x86_64   
OS: Linux   
Whiteboard: abrt_hash:565693110765629e9050d67c3ea4d5d48a4e5125
Fixed In Version: gdb-7.0.1-34.fc12 Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2010-03-16 00:50:26 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Attachments:
Description Flags
File: backtrace
none
Source file with an inline function none

Description Jerry James 2010-01-05 17:09:01 UTC 
abrt 1.0.0 detected a crash.

How to reproduce
-----
1. Compile XEmacs with these flags: -march=core2 -pipe -D_FORTIFY_SOURCE=2 -feliminate-dwarf2-dups -fno-strict-aliasing -O2 -g3
2. Start gdb with "gdb src/xemacs"
3. At the gdb prompt, type "run"

Comment: I'm guessing that either -feliminate-dwarf2-dups or -g3 produced dwarf2 output that gdb didn't expect.  I've used the former for many years, because it drastically reduces the size of the debug information.  I'll try dropping one or both of these and see if the gdb crash goes away.
Attached file: backtrace
cmdline: gdb ./xemacs
component: gdb
executable: /usr/bin/gdb
kernel: 2.6.31.9-174.fc12.x86_64
package: gdb-7.0.1-19.fc12
rating: 4
reason: Process was terminated by signal 11
Comment 1 Jerry James 2010-01-05 17:09:03 UTC 
Created attachment 381798 [details]
File: backtrace
Comment 2 Jerry James 2010-01-05 17:15:14 UTC 
Sure enough, without -feliminate-dwarf2-dups, gdb behaves normally.
Comment 3 Tom Tromey 2010-02-25 20:50:59 UTC 
My first suspicion is that there is an underlying gcc bug.
What version of gcc did you use?
(gdb ideally should not crash on invalid dwarf, but my
understanding is that some checking is omitted because it
would slow down startup too much.)
Comment 4 Jerry James 2010-02-25 21:11:01 UTC 
I apply updates aggressively, so there's no telling. :-)  However, I just tried it again with the same results.

gcc-4.4.3-4.fc12.x86_64
gdb-7.0.1-31.fc12.x86_64

I wonder if this has something to do with XEmacs' aggressive use of inline functions.  I'm going to try a little test and see if I can trigger the bug.
Comment 5 Jerry James 2010-02-25 21:15:36 UTC 
Created attachment 396395 [details]
Source file with an inline function

If I compile this source file with -feliminate-dwarf2-dups, then on starting up GDB, it warns:

Reading symbols from /tmp/test/inline-test...done.
DW_FORM_strp pointing outside of .debug_str section [in module /tmp/test/inline-test]

but doesn't if I omit the -feliminate-dwarf2-dups flag.  I have no idea if this has anything to do with the crash.  Maybe I need to define LOTS of inline functions....
Comment 6 Tom Tromey 2010-03-04 21:19:30 UTC 
Thanks for the test case.
This is definitely a gdb bug, not a gcc bug.
The problem is that DWARF 3 specified a change in the
size of DW_FORM_ref_addr, and gdb did not implement this change.
I'm working on it.
Comment 7 Tom Tromey 2010-03-04 22:02:26 UTC 
I pushed the fix into upstream gdb.
Jan, this should go into our releases as well.
http://sourceware.org/ml/gdb-patches/2010-03/msg00219.html
Comment 8 Jerry James 2010-03-04 22:07:46 UTC 
Thanks for tracking that down, Tom.
Comment 9 Fedora Update System 2010-03-12 13:58:40 UTC 
gdb-7.0.1-34.fc12 has been submitted as an update for Fedora 12.
http://admin.fedoraproject.org/updates/gdb-7.0.1-34.fc12
Comment 10 Fedora Update System 2010-03-16 00:50:05 UTC 
gdb-7.0.1-34.fc12 has been pushed to the Fedora 12 stable repository.  If problems still persist, please make note of it in this bug report.