This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
Update on i686 bootstrap failure, please revert the patch
- To: gcc-regression at gcc dot gnu dot org, gcc-bugs at gcc dot gnu dot org,gcc-patches at gcc dot gnu dot org
- Subject: Update on i686 bootstrap failure, please revert the patch
- From: Daniel Berlin <dan at cgsoftware dot com>
- Date: Fri, 03 Aug 2001 22:05:40 -0400
- Cc: dan at cgsoftware dot com, rth at localhost dot localdomain
- References: <200108040006.f7406SK18849@maat.cygnus.com>
[Copied to gcc patches to get attention of someone who can do the reversion]
Just an update on the inability to bootstrap on i686.
I'm still working on it, but the bug isn't immediately obvious to me.
It also bootstraps okay on other platforms, and has no regressions.
Its also fun in that if you bootstrap with 3.0 as the compiler for the
first stage compiler, you get a failure a lot sooner than if you bootstrap with
2.95.4 as the compiler for the first stage. In fact, with 2.95.4, it's
not until it actually goes to run the resulting cc1/cpp0 that
it fails.
However, i'm sure there is a bug somewhere in store motion, and it's
going to take a while to find it.
I'm not feeling well at all right now, and am probably going to be
down for the weekend, so can someone please revert the patch for me (I
f I do it, i'm sure i'll screw it up in my condition) until I can find
the bug causing x86 bootstrap to fail?
You can actually comment out the call to store_motion if you don't
want to revert the whole thing. I've bootstrapped and regression
tested #if 0'ing the call out, on x86 and powerpc, to make sure it
wasn't the part of the changes that happen to improve load motion
screwing us up. Whatever is easiest for someone is fine by me.
Thanks,
Dan