On 21 November 2004 12:45, Sven Panne wrote:

> Sven Panne wrote:
>> panne       2004/11/21 04:13:29 PST
>> 
>>   Modified files:
>>     ghc/utils/ghc-pkg    Makefile
>>   Log:
>>   Fixed include paths. Thinking about the convoluted dependencies
>>   between ghc, ghc-pkg and libraries really makes me dizzy... Some
>>   cleanup and simplification is needed here *urgently*, this stuff
>> is currently beyond my grasp. 
>> 
>>   Revision  Changes    Path
>>   1.32      +2 -1      fptools/ghc/utils/ghc-pkg/Makefile
>> _______________________________________________
>> Cvs-ghc mailing list
>> [EMAIL PROTECTED]
>> http://www.haskell.org/mailman/listinfo/cvs-ghc
> 
> Hmmm, this fix is wrong/incomplete. Now I have to add "-package
> Cabal" to link ghc-pkg when building the HEAD with itself, but how
> can this ever bootstrap with older GHCs?? Could somebody be so kind
> and write a few lines explaining the intricate dependencies and
> concepts here (ghc/ghc-pkg/libcompat/package Cabal)? I just don't get
> it... 

The current situation is explained in ghc/lib/compat/README.  It is
difficient in a few ways, and I'd appreciate any ideas you (or others)
have on how to improve the situation:

  - libghccompat.a has some dependencies, notably some external
    libraries on Windows, and -package Cabal if GHC >= 6.3.  Since it
    isn't a package, these dependencies have to be added to SRC_LD_OPTS
    explicitly when using libghccompat.a.

  - mkdependHS needs to be told not to include modules from
ghc/lib/compat
    when generating dependencies.  mkdependHS doesn't have an
--exclude-dir
    option, only --exclude-module.  Again, if libghccompat.a were a
    package, this wouldn't be a problem.

Making libghccompat a package seems the obvious way forward.  However,
then we need a local package.conf file, and we have to deal with
different package config formats (6.4+ will have a new package.conf
syntax).

Cheers,
        Simon
_______________________________________________
Cvs-ghc mailing list
[EMAIL PROTECTED]
http://www.haskell.org/mailman/listinfo/cvs-ghc

Reply via email to