6.4.3 and threaded RTS problems

[Simon Marlow]

David Kirkman wrote:
> I ran the testsuite against the current HEAD (as of saturday afternoon)
> and had far fewer failures.  Is the last week's work supposed to have
> fixed so many test cases?  Or are different machines giving different
> results?  The tests that resulted in unexpected failures for greg were
> all run on my machine, but many of them worked on my setup.

We have put some work into cleaning up the test failures, currently the 
x86_64/Linux build (my main dev environment) is at 28 failures, most of which 
are probably not real bugs.


> Unexpected passes:
>   barton-mangler-bug(optasm)
>   cholewo-eval(optasm,profasm)

these two are susceptible to floating-point differences, I wouldn't worry about 
them.

> Unexpected failures:
>   TH_repPatSig(normal)
>   barton-mangler-bug(opt)
>   conc019(opt)
>   conc024(normal)
>   conc039(threaded1)
>   conc056(threaded1)
>   ffi009(ghci)
>   ffi016(normal,prof)

the conc and ffi failures are worrying.

>   gadt7(normal)
>   galois_raytrace(opt,prof)
>   maessen_hashtab(normal,opt,optasm,prof,profasm,ghci,threaded1)
>   rnfail026(normal)
>   signals002(ghci)
>   tcfail068(normal)
>   tcfail071(normal)
>   tcfail090(normal)
>   tcfail099(normal)
>   tcfail103(normal)
>   tcfail115(normal)
>   tcfail140(normal)
>   tcrun021(normal,opt,optasm,prof,profasm,ghci,threaded1)

these are all known, except for galoiis_raytrace which is probably a 
floating-point difference.

>   utf8_002(normal)
>   utf8_003(normal)
>   utf8_004(normal)
>   utf8_005(normal)

I believe you can make these go away by installing the latest Alex from darcs. 
It's nothing to worry about, I made a small tweak to Alex that improved the 
error messages from GHC's lexer.

Cheers,
	Simon