Re: [Gm2] latest regression tests

[Michael Lambert]

Hi Gaius,

I'm still seeing

# of expected passes            5964
# of unexpected failures        60
# of unresolved testcases       12

I think this is the culprit in the six extra failures:

FAIL: gm2/pim/options/optimize/run/pass/testadd.mod compilation,  -g
Link libraries are:
ld: in /var/folders/5r/5rG8VggWHx4wBa0Sofcs6U+++TI/-Tmp-// 
cciNl0KL.a(addition.o)
, not a valid mach-o object file
collect2: ld returned 1 exit status

(I'll forward the entire output to Gaius)

A couple of other notes for Mac OS X/Intel:

o Paranoid passes.

o On Mac/Intel the assembler has changed from vanilla gcc 4.1.2, so  
the following patch is needed for the diffs to work (the generated  
code is fine without them):

--- gcc-4.1.2/gcc/config/i386/i386.c.sav	2006-11-17 02:01:22.000000000  
-0500
+++ gcc-4.1.2/gcc/config/i386/i386.c    2008-04-22 12:59:31.000000000 -0400
@@ -16666,7 +16666,7 @@
     {
       fprintf (file, "\tcall LPC$%d\nLPC$%d:\tpopl %%eax\n", label,  
label);
       fprintf (file, "\tmovl %s-LPC$%d(%%eax),%%edx\n",  
lazy_ptr_name, label);
-      fprintf (file, "\tjmp %%edx\n");
+      fprintf (file, "\tjmp *%%edx\n");
     }
   else
     fprintf (file, "\tjmp *%s\n", lazy_ptr_name);

Since there are conditionals around the above code anyway, I think it  
should probably be put into the "official" gm2/patches/gcc/4.1.2  
directory.

Michael

On 30 Apr 2008, at 09:15, Gaius Mulley wrote:

> Hi Michael,
>
> thanks for the report - I still get the above arrayhuge.mod failure -
> probably not macos specific.  Looks like the two systems are at least
> running the same tests :-)
>
> > I don't quite remember where I need to make changes (I vaguely recall
> > that you once sent me a piece of code to determine the appropriate
> > values).
>
> I think this is due to a bug in cast - which I believe I've fixed.
> Certainly on i[456]86 GNU/Linux it passes.  Would it be possible for  
> you
> to run this again with the latest snapshot?
>
> I'm currently getting these results from a 32 bit GNU/Linux (Debian
> Etch) system:
>
> # of expected passes            5970
> # of unexpected failures        54
> # of unresolved testcases       12
>
> Thanks,
> Gaius