[Bug-dejagnu] Re: Ignore some warnings on IRIX 6 O32

[Eric Christopher]

> The first set of changes is trivial (i.e. ignore warnings from all
> frontends, not only cc1 and cc1plus).  The other ones are more interesting
> and might point to problems in GCC O32 code generation:
> 
> * I get this one in many places (only with -mips1, it seems), e.g. in
>   g77.f-torture/compile/19990826-3.f,  -O1:
> 
> as1: Warning: /var/tmp//cc5ooMCa.s, line 1: shouldn't put mult/div in branch 
> delay slot (R10K bug)
> 
>   The line number is wrong since as mishandles 
> 
>         .file   1 
> "/vol/gnu/src/gcc/gcc/gcc/testsuite/g77.f-torture/compile/1999
> 0826-3.f"
> 
> * This one happens e.g. in gcc.c-torture/compile/920501-4.c,  -O1:
> 
> as1: Warning: /var/tmp//ccPJlzBd.s, line 1: DIV in branch delay slot may not 
> work on a R4000 chip
> 

These two are interesting. I don't know that we want to work around
every processor bug by default, but if it is a processor bug that we
know about and don't work around we should definitely fix that.

> * gas 2.11.2 warns
> 
> /var/tmp//cctWmzAa.s: Assembler messages:
> /var/tmp//cctWmzAa.s:160: Warning: Branch bltu is always false (nop)
> 
>   e.g. in g++.other/reload1.C (test for excess errors).  I need to ignore
>   both the intro line and the specific warning.  In gas 2.12, this warning
>   is off by default, but could be enabled with -n.

This is on my list of things to fix. Non-conditional branches should
probably be avoided and fixed up.

The others look just fine.

-eric

-- 
I will not carve gods