On Wed, Nov 16, 2011 at 04:16:54PM -0600, Bob Tracy wrote:
> The subject release appeared less than a day ago, so it seemed like a
> good time to try out gcc/g++-4.6. Will try a non-optimized build
> without the "nsThreadUtils*" patches to see how that goes. When I last
> tried such a build (version 4.0.1), I got segfaults until I put the
> "nsThreadUtils*" patches back in :-(. Still don't understand *why*
> those patches (which are essentially a downgrade to older code) should
> have any effect: by inspection, the new code appears to be equivalent.
> Obviously, the compiler thinks otherwise :-(, which is why I'm trying
> the current build with the new compiler and without those patches.
Didn't want to spam the list with updates until I actually had something
good to report. Michael got the sporadic news from the battlefield :-).
firefox-8.0.1 is built. Between 4.0.1 and 8.0.1, enough has changed
that the build times on my PWS 433au have increased from 8 hours to just
under 2 days. Using gcc/g++-4.6 was successful: no optimization. The
"nsThreadUtils*" patches were required. Michael reports success with
"-O1" and *no* "nsThreadutils*" patches. Default optimization (-O2)
fails regardless of whether the "nsThreadUtils*" patches are used.
Looks like we may have a place to begin looking for the optimizer bug.
The ES45 machines have arrived! An "iceweasel" build should be
available in the Alpha "unstable" repository in a few days.