XFAIL for ia64 990208-1

[Jim Wilson]

In article < 200109191925.MAA26709@rtl.cygnus.com > you write:
>Unless there is disagreement, I will check in the following file
>into the mainline and gcc3.0 branch.

You can't check in patches just because no one objected soon enough.
You can only check in patches if they are approved.

I found your earlier thread in July by searching gcc-patches for 990208-1.
http://gcc.gnu.org/ml/gcc-patches/2001-07/msg01669.html
I also found an earlier thread in May from Richard Henderson containing
the if-convert patch that broke this testcase, and an attempt to modify
the testcase to keep it working.
http://gcc.gnu.org/ml/gcc-patches/2000-05/msg00723.html
Considering both threads, it is clear that we can't make this testcase
work anymore for targets with COND_EXEC support at -O3.

An alternative fix might be making -fno-inline-functions a required option
for this testcase, which should work if we can make sure that it always comes
after the -O3.  I don't know if we can do that, though.  Simplying marking
it XFAIL for IA-64 at -O3 seems reasonable for now, and fixes an IA-64
testsuite blemish.

The patch is OK for both mainline and the gcc 3.0 branch.

Jim