This is the mail archive of the
mailing list for the GCC project.
Re: m68k bootstrapping broken
On Tue, Jan 13, 2004 at 04:57:34PM +0100, Andreas Schwab wrote:
> Gunther Nikl <email@example.com> writes:
> > On Tue, Jan 13, 2004 at 03:37:35PM +0100, Andreas Schwab wrote:
> >> Gunther Nikl <firstname.lastname@example.org> writes:
> >> > I just built a native compiler for m68k-amigaos with a cross-compiler
> >> > that had the updated patch applied. No objects in gcc/ except m68k.o
> >> > differed. Is that a good or bad sign?
> >> Can you please find out where exactly they differ?
> > I thought that would be obvious: because of your patch (at least I think
> > so). I could build a version with the patch applied by an unpatched cross
> > compiler to compare the result with objects from the build with the patched
> > cross-compiler.
> I thought you were talking about stage difference during bootstrap. Is
> m68k-amigaos self-hosting?
AFAICT, it is _but_ I never bootstrapped any 3.x (starting with 3.2.2)
because it would probably take very much time and I don't have enough
patience. All I did with 3.2.2 was to built a stage1 compiler without
> > Note: the "differed" part in my original mail refered to a compare with an
> > unpatched native compiler built with an unpatched cross-compiler. Somehow
> > I expected other differences besides m68k.o because this patch is supposed
> > to fix bugs. But maybe for my target the bug wasn't triggered by the GCC
> > sources with the flags I used for compilation.
> The only known miscompilation is in insn-recog.o, and only when building
> with -O2 (which implies -funit-at-a-time).
In that case lshrdi.o causes an ICE, right? If insn-recog.o is
miscompiled, then it should change when compiled by a patched
cross-compiler, shouldn't it?