[nightly] 11-Oct-2004 build of STABLE on Darwin 7.5.0 powerpc(Stealth.local)

[Wolfgang Thaller]

>>    enum01(normal)
>>    enum01(opt)
>>    enum01(optasm)
>>    enum01(prof)
>>    enum01(ghci)
>>    enum01(threaded)

Something to do with the absolutely evil preprocessor black magic that 
this testcase does. gcc -E, gcc -traditional -E, gcc3 -E and gcc3 
-traditional -E all yield completely different results on the 
testcase's source code (gcc is [Apple's version of] gcc 3.3, gcc3 is 
[Apple's version of] gcc 3.1).


>>    memo001(ghci)
>>    memo002(ghci)

ghc-6.2.2: can't load .so/.DLL for: readline (dlcompat: file 
"libreadline.dylib" not found)

Package util requires readline. There's no dynamic version of readline 
on the system, but a static version, so everything except ghci works. 
No problem here.

>>    arith016(normal)
>>    arith016(optasm)
>>    arith016(threaded)

The old NCG treats some floating point comparisons involving NaNs. It's 
fixed in the HEAD, so I won't fix it in the old NCG.

>>    tcrun021(ghci)

Imports FiniteMap from the old hslibs package data, which depends on 
util, which depends on readline. Same problem as for the memo tests 
above, no dynamic readline is installed.

Most of the time, we don't want to install a dynamic version of 
readline on a ghc build machine; we don't want GHC itself to depend on 
a dynamic version of readline as installing dynamic libraries on Mac OS 
X is a task that requires some UNIX knowledge (which many 
first-year-students who need GHC for one of their classes don't have).

So there are no show-stoppers, only I [or someone else who has a Mac 
and write access to the repository ;-) ] should do some cleaning-up in 
the testsuite, so that those things don't show up as unexpected 
failures.

Cheers,

Wolfgang