RE: "Hello World" program core dumps with GHC 4.08.1 on HPUX 10.2 0

[BENNETT,ANDY (HP-Unitedkingdom,ex1)]

Hi Simon,

I'm getting ever closer... I can now get it to produce a compiler that
doesn't core dump. I wonder if you could have a look in your notes to see if
you have any information about a couple of points. Firstly, a while back you
mentioned that you had disabled "long long" support when you built 4.08.1.
Could you let me know what you did in case I need to do the same? Secondly,
the modules RtsStartup.c and Main.c both contain references to haskell
functions __init_Prelude and stg_init_ret (?) and __init_PrelMain
respectively, that need their pointers mangling. Since they are plain C, ghc
doesn't call the mangler. Currently, I've hand mangled the references. Can
you remember if there are any other references like this and what you did
about them when you built 4.08.1?

It would be nice to get it into a state where it can truely build itself
automatically on HP-UX without the need for a degree in compiler
engineering, in the same way as I assume it does on the other platforms.

Regards,
Andy.

> -----Original Message-----
> From: Simon Marlow [mailto:[EMAIL PROTECTED]]
> Sent: 04 April 2001 02:51
> To: BENNETT,ANDY (HP-Unitedkingdom,ex1);
> [EMAIL PROTECTED]
> Subject: RE: "Hello World" program core dumps with GHC 4.08.1 on HPUX
> 10.2 0
> 
> 
> Hi Andy,
> 
> > I think I've made most of the changes in the mangler to get 
> > it to work with
> > gcc 2.95.2 on HP-UX (only about 6 lines). One question 
> > however, the *_info
> > is inverted and placed in the code section I presume so that 
> > its address is
> > identical to the entry point for an associated function. Is 
> > it only for the
> > *_entry of a function that this adjacency matters? I'm 
> > interested to know as
> > I'd like to know what potentially needs to be in the same 
> > subspace as the
> > info table, as the linker reserves the right to reorder subspaces.
> 
> Info tables (*_info) need to be right up against their corresponding
> entry code (*_entry) for closures, or return code (*_ret) for return
> addresses.  Vector tables (*_vtbl) can float around in 
> relation to their
> code (*_alt), but they have to be inverted too.
> 
> > I've got a couple of small programs working, and without the 
> > need for my
> > original procedure label workaround. Once you let me know the 
> > above, I'll
> > see if I can get GHC to build itself, and if that works, I'll 
> > let you know
> > my changes to the mangler.
> 
> Sounds great!  I think we owe you a beer or two :)
> 
> Cheers,
>       Simon
> 

_______________________________________________
Glasgow-haskell-bugs mailing list
[EMAIL PROTECTED]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs