mips64 and gmp observation

[Matt Chapman]

Hi Don,

The same is true on IA64, if you look at the configure script there's
a little hack of mine...

if test "$HaveLibGmp" = "NO"; then
if test "$HostArch_CPP" = "ia64"; then
AC_MSG_ERROR([You need to install libgmp (the in-tree version does not work on IA64).])
fi;
fi;

It's probably broken for 64-bit architectures in general.

Matt


On Sun, Oct 26, 2003 at 07:19:58PM +1100, Donald Bruce Stewart wrote:
> Hey all,
> 
> Some mips64 and gmp observations:
> 
> The normal way, unregisterised, with the in-tree gmp, v3.1.1:
> 
>    OVERALL SUMMARY for test run started at Fri Oct 24 17:17:56 PDT 2003
>         1070 total tests, which gave rise to
>         1070 test cases, of which
>            0 caused framework failures
>           29 were skipped
> 
>          889 expected passes
>           10 expected failures
>            0 unexpected passes
>          142 unexpected failures
>         ^^^^^
> When I recompiled using the Irix-provided libgmp, also v3.1.1, 
> the results were:
> 
>    OVERALL SUMMARY for test run started at Sat Oct 25 20:13:03 PDT 2003
>         1070 total tests, which gave rise to    
>         1070 test cases, of which        
>            0 caused framework failures   
>           29 were skipped                
>                                        
>          932 expected passes
>           10 expected failures           
>            0 unexpected passes           
>           99 unexpected failures
>         ^^^^^
> 
> 43 tests, involving floats or double, passed with the external libgmp.
> The brokeness exhibited itself as segfaults, or showing values as 0.0 or
> 0.0e0 instead of the correct result.
> 
> These tests that were cured by using an external gmp:
> 
>     arr0{05,06,13,16}
>     num0{01,07,08,11}
>     cg0{14,18,19,24,26,28,34,35,42}
>     arith0{01,04,06-08,12-14,16-17}
> 
>     CPUTime001 gshow gzip paradise strings
>     timeexts001 rand001 dynamic001 mdo001
> 
>     andy_cherry barton-mangler-bug cholewo-eval
>     fun_insts jl_defaults lennart_range thurston-modular-arith
> 
>     tcrun027
> 
> So mips64 users should definitely use the external gmp provided with
> Irix, as it must be better configured than ours for the time being.
> 
> And, I guess all porters should take note that even though a port
> builds, and can then rebuild ghc itself, it may still be broken in ways
> only the testsuite will find.
> 
> Cheers,
>     Don
> _______________________________________________
> Glasgow-haskell-users mailing list
> Glasgow-haskell-users at haskell.org
> http://www.haskell.org/mailman/listinfo/glasgow-haskell-users