Re: egcs-1.1 prerelease vs egcs-1.0.3a on stock Red Hat 5.1 Alpha

[Richard Henderson <rth at cygnus dot com>]

On Wed, Jul 22, 1998 at 10:50:53PM +0200, Toon Moene wrote:
> FAIL: gcc.c-torture/execute/980506-2.c execution,  -O2

This is a general aliasing problem.  It fails on quite a
lot of machines.

> FAIL: gcc.c-torture/execute/980526-1.c execution,  -O2  

This should be examined.

> FAIL: gcc.c-torture/execute/complex-5.c execution,  -O0

SCmode will continue to fail until complex argument passing
is reworked in the middle end.

> FAIL: gcc.c-torture/execute/ieee/mzero2.c execution,  -O0

In /usr/local/share/dejagnu/baseboards/unix.exp, put

if { [istarget "alpha*-*-*"] } {
        set_board_info ieee_multilib_flags "-mieee";
}

With this change, there should be no IEEE failures.

> XPASS: g++.benjamin/p12475.C (test for excess errors)
> XPASS: g++.brendan/array1.C overflow in array dimension.* , (test  
> for errors, line 6)

Whee.  Gotta make those constants bigger to break 64-bit machines.  ;-)

As for the rest of the C++ bits, I'm used to seeing about a half
dozen fail.  I'll look again at the rest.


r~