Unbreaking news (was: Re: Trunk broken for MMIX, CRIS, i960, m32r, (v850, ARM))

[Hans-Peter Nilsson]

On Sat, 5 Jul 2003, Hans-Peter Nilsson wrote:

> On Sat, 5 Jul 2003, Andreas Jaeger wrote:
> > Andrew Pinski <pinskia@physics.uc.edu> writes:
> > > On Saturday, Jul 5, 2003, at 00:20 US/Eastern, Andreas Jaeger wrote:
> > >> Andrew, did you test Zdenek's patch (Subject: Re: cfg_cleanup and loop
> > >> discovery in cfg_layout mode) for cfglayout?  Does it fix the problem
> > >> for you?
> > > No I have not because there was two different bootstrap failures (both
> > > caused by Zack's patches)
> > > in between then and now, I will try it soon (one still is not fixed by
> > > I have submitted a patch
> > > to fix it).
> > Ok, I'll test Zdenek's patch now also on x86_64-linux-gnu (already
> > passed the critical point there) and on i686-linux-gnu and let's see
> > what it does - I would appreciate some review of Zdenek's patch to get
> > it in asap.
>
> I actually noticed that patch
> (<URL:http://gcc.gnu.org/ml/gcc-patches/2003-07/msg00457.html>)
> but confused it with another patch by Zdenek that was committed,
> so I didn't look closer.
>
> I'm testing to see it it unbreaks all targets I mentioned.  It
> helps for mmix-knuth-mmixware at least; it now builds (all
> supported languages except Ada and treelang).

Though I see these failures which are regressions from last
successful build (LAST_UPDATED Thu Jul  3 10:03:12 GMT 2003):
FAIL: gcc.c-torture/compile/20000728-1.c (test for excess errors)
FAIL: gcc.c-torture/compile/920729-1.c (test for excess errors)
FAIL: gcc.c-torture/compile/930503-1.c (test for excess errors)
FAIL: gcc.c-torture/compile/940611-1.c (test for excess errors)
FAIL: gcc.c-torture/compile/940718-1.c (test for excess errors)
FAIL: gcc.c-torture/compile/990617-1.c (test for excess errors)
FAIL: gcc.c-torture/compile/990801-1.c (test for excess errors)
FAIL: gcc.c-torture/compile/991213-2.c (test for excess errors)
FAIL: gcc.c-torture/compile/goto-1.c (test for excess errors)
FAIL: gcc.c-torture/compile/iftrap-1.c (test for excess errors)

These tests are all using the compiler flags "-O3
-fomit-frame-pointer -funroll-all-loops -finline-functions"
(they used to be visible in the summary, but that was
unfortunately lost with the dg changes to these tests, hi Mark
[you asked for dg fallout]) and the dying messages are on the
form (beware cut-n-pastage):

x/gcc/testsuite/gcc.c-torture/compile/20000728-1.c: In function `foo':
x/gcc/testsuite/gcc.c-torture/compile/20000728-1.c:15: internal \
compiler error: virtual array dominance info[7]: element \
1074822522 out of bounds in set_immediate_dominator, at dominance.c:621

(You don't need the mmix simulator to repeat this; just build
cc1 for mmix-knuth-mmixware.)

> I'm cfgloop ignorant and can only comment on minor style issues;
> the comment on the abort () line should be on a separate line,
> and empty lines before and after the
> !   if (!insn)
> !     abort ();  /* No basic blocks at all?  */
> lines wouldn't hurt.
>
> Thanks to Zdenek and Andreas for keeping an eye on things.
>
> brgds, H-P