Subject: pkg/35946: www/firefox2 needs -fno-strict-aliasing
To: None <pkg-manager@netbsd.org, gnats-admin@netbsd.org,>
From: None <blair.sadewitz@gmail.com>
List: pkgsrc-bugs
Date: 03/07/2007 21:10:01
>Number:         35946
>Category:       pkg
>Synopsis:       www/firefox2 needs -fno-strict-aliasing
>Confidential:   no
>Severity:       non-critical
>Priority:       medium
>Responsible:    pkg-manager
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed Mar 07 21:10:00 +0000 2007
>Originator:     Blair Sadewitz
>Release:        4.99.13 amd64
>Organization:
>Environment:
NetBSD dexamyl 4.99.13 NetBSD 4.99.13 (DEXAMYL) #0: Wed Mar  7 15:14:43 EST 2007
  blair@dexamyl:/home/blair/fresh-build/src/sys/arch/amd64/compile/DEXAMYL amd64

>Description:
Building this without -fno-strict-aliasing yields many, many warnings about type-punning breaking strict aliasing rules.  I usually build it myself with -fno-strict-aliasing, but this time (on i386, not this machine, but it happens on amd64 too, obviously) I let the build proceed.

When I started firefox, it immediately segfaulted (I think it was "Memory error").  I cannot say for certain that this was the cause, but IMHO it couldn't hurt to comply with gcc's wishes, as the warnings are plentiful.  In some instances, the could could break the rules, but in others it most definitely does.

I do not know if this applies to www/firefox.
>How-To-Repeat:
Build www/firefox2 without -fno-strict-aliasing.
>Fix:
Add -fno-strict-aliasing to CFLAGS in www/firefox2/Makefile or www/firefox/Makefile.common.