This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: PATCH: [3.3 branch] IA64 bootstrap failure
On Tue, Jul 08, 2003 at 10:21:30PM -0700, Zack Weinberg wrote:
> "H. J. Lu" <hjl@lucon.org> writes:
>
> > On Tue, Jul 08, 2003 at 10:02:13PM -0700, Zack Weinberg wrote:
> >> "H. J. Lu" <hjl@lucon.org> writes:
> >>
> >> > On Tue, Jul 08, 2003 at 11:19:13PM -0400, Hans-Peter Nilsson wrote:
> >> >> On Tue, 8 Jul 2003, H. J. Lu wrote:
> >> >> > I am testing this patch now.
> >> >>
> >> >> +int x [2] =
> >> >> +{
> >> >> + __builtin_constant_p (10) ? 10 : -1,
> >> >>
> >> >> All other issues aside, this part doesn't cause the intended
> >> >> failure if __builtin_constant_p falsely returns false; it'll
> >> >> just set that element to -1. You want two arrays or a
> >> >> two-dimensional array, with those (negative or variable-indexed)
> >> >> expressions in the []-parts.
> >> >>
> >> >
> >> > This patch should be conservative enough.
> >>
> >> What part of "FIX THE REAL BUG" don't you understand?
> >>
> >
> > For that particular problem, the file is NOT miscompiled by the stage2
> > compiler. It is just different from the one compiled by the stage1
> > compiler. Since stage1 and stage2 have different source files, see
> > the diffs enclosed here, it doesn't make much senses to compare outputs
> > from those 2 compilers. Unless there is a difference between stage1
> > and stage2 of the same source files, such a comparison doesn't mean
> > much.
>
> Why are you comparing stage1 and stage2, when the proper comparison is
> stage2 and stage3, which will have the same source files?
Let me get this straight:
1. .o files under gcc/stage1/ are compiled by the bootstrap compiler.
2. .o files under gcc/stage2/ are compiled by the stage1 compiler.
3. .o files under gcc/ are compiler by the stage2 compiler.
Since we compare .o files under gcc/stage2/ against .o files under gcc/,
why shouldn't I compare .c files under gcc/stage1/ against gcc/stage2/
for differences between stage1 and stage2 compiler sources.
H.J.
H.J.