On Wed, Nov 26, 2008 at 08:43:18PM -0600, Steve M. Robbins wrote: > On Wed, Nov 26, 2008 at 11:34:31AM -0800, Steve Langasek wrote: > > On Wed, Nov 26, 2008 at 02:01:53AM -0600, Steve M. Robbins wrote:
> > > The patch bjam-no-strict-aliasing only applies to the second build > > > (using the bootstrap bjam). We need to supply > > > CFLAGS=-fno-strict-aliasing on the first bjam build because otherwise > > > the first bjam segfaults on some architectures, e.g. alpha [1]. > > That's true if not specifying the gcc profile when building; if you do (as > > in the Ubuntu patch), then it applies equally well to both builds. > Examining build.sh, I don't see any evidence of that. Hmm, you're right; it turns out that this approach works only because the gcc profile is not using -O2 by default, and the aliasing bug only manifests under -O2. So this proposed patch isn't actually any more correct than the current behavior, sorry about that; I'll drop that change out of the Ubuntu package in the next upload. > > I know > > this to be true because the Ubuntu patch was added to address the exact same > > issue, but I've also tested that the Ubuntu package builds exactly as > > intended on alpha, too. > The crashes appear to happen at unpredictable times. I suspect you > merely got lucky. Hrm, no, it's not very unpredictable at all - if you build with -O2 and without -fno-strict-aliasing on alpha, it segfaults 100% of the time. -- Steve Langasek Give me a lever long enough and a Free OS Debian Developer to set it on, and I can move the world. Ubuntu Developer http://www.debian.org/ [EMAIL PROTECTED] [EMAIL PROTECTED] -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
