Bug 202055

Summary:	sys-libs/glibc: cfi frame pointers in clone triggers bugs in gcc unwinder
Product:	Gentoo Linux	Reporter:	Hanno Zysik (geki) <h.mth>
Component:	[OLD] Core system	Assignee:	Gentoo Toolchain Maintainers <toolchain>
Status:	RESOLVED FIXED
Severity:	normal
Priority:	High
Version:	2007.0
Hardware:	All
OS:	Linux
URL:	http://sourceware.org/bugzilla/show_bug.cgi?id=4781
Whiteboard:
Package list:		Runtime testing required:	---
Bug Depends on:
Bug Blocks:	346373
Attachments:	fix cpu/memory kill

Description Hanno Zysik (geki) 2007-12-12 16:59:14 UTC

There is a regression(?) in glibc-2.7 which was fixed in 2.6.1 afai understand.

There is a diff in archlinux cvs which fixes this. May be related to bug 199009?

Reference:
http://sourceware.org/ml/binutils/2007-11/msg00069.html

Diff:
http://cvs.archlinux.org/cgi-bin/viewcvs.cgi/base/glibc/Attic/fix_memory.diff?rev=1.1&cvsroot=Core&only_with_tag=CURRENT&content-type=text/vnd.viewcvs-markup

Comment 1 Hanno Zysik (geki) 2007-12-12 17:00:24 UTC

Created attachment 138329 [details, diff]
fix cpu/memory kill

Comment 2 SpanKY gentoo-dev

2007-12-30 11:46:52 UTC

this is not a glibc bug ... it's actually a bug in gcc.  read the full thread and you'll see that the cfi stuff was reverted only to workaround the issue, not fix it.

Comment 3 Hanno Zysik (geki) 2007-12-30 15:38:17 UTC

Yes, I knew, sorry. I got a bit carried away; realized after commit. :(

Comment 4 SpanKY gentoo-dev

2008-03-23 15:13:17 UTC

worked around in glibc-2.7-r2 ... will drop once newer gets into ~arch