Re: .hi problem with existentially quantified type variables

[Manuel M. T. Chakravarty]

Sven Panne <[EMAIL PROTECTED]> wrote,

> [ Aaaargl, 2nd try: The ancient mailer at Glasgow didn't like my
>    MIME-attachment.  >:-( ]

Why does it care at all about attachments?

> > as I am still in the middle of compiling it. It seems as if - at
> > least for Linux - the heap size is too small for many of the library
> > modules.  Sven, did you have the same problem?  (Or can it
> > be a problem that I compiled the compiler proper with 3.02?)
> 
> I'm always compiling ghc with the latest installed version, e.g.
> Friday's ghc is compiled with Thursday's. Additionally, hsc is compiled
> with -O, so I can't comment on non-optimised builds. My RPM patch file
> is appended. 

Maybe -O makes the difference.

> But on Linux you can always use the pre-built RPMS on
> 
>    ftp://ftp.informatik.uni-muenchen.de/pub/local/pms/

Thanks, but since my 62-tuple patch never made it in the
main distribution, I have to do my own RPM anyway.  (And it
is nice to be able to tweak it...)

> BTW, I'd like to get some feedback from RedHat users: Byron Hale
> reported some problems with the Hugs RPM (the dynamic linker tells
> someting about an undefined symbol __register_frame_info).
> Success/Failure stories are highly welcome.

Sorry, but I can't comment on RedHat.  I am using yet
another distribution (albeit one very similar to RedHat):
TurboLinux - the only distribution that has a Japanese
version (Japanese e-mail messages don't look very nice on an 
English system ;-)

Re egcs: TurboLinux also uses egcs, but I replaced it with
gcc, after having some trouble (in particular with
GHC...although, this is supposed to be resolved now).

Cheers,

Manuel