References to the Rts in unregisterised first build of ghc 6.4

[Simon Marlow]

On 27 June 2005 17:53, Michiel Buddingh' wrote:

> On Mon, Jun 27, 2005 at 01:00:01PM +0100, Simon Marlow wrote:
>> On 26 June 2005 18:28, Michiel Buddingh' wrote:
>>> I was able to generate a set of .hc files on a linux-i386 host, but
>>> (as was expected) the rtc failed to build.
>> 
>> Where exactly did the RTS build fail?  Can you post the error
>> message? We do need the .hc files generated from the RTS .cmm source
>> files. 
> 
> Adjustor.c:106:27: machine/pal.h: No such file or directory
> Adjustor.c: In function createAdjustor':
> Adjustor.c:403: error: PAL_imb' undeclared (first use in this
> function) Adjustor.c:403: error: (Each undeclared identifier is
> reported only once Adjustor.c:403: error: for each function it
> appears in.) 
> 
> I didn't take a look at this before, but I suspect machine/pal.h
> might even be specific to OSF/1.

Ok, I suggest you do a 'make -k' in ghc/rts.  The .cmm files should at
least generate .hc files, that you can take to the target machine.

> Other alterations (in reverse order)
> 
> Edited out the -mieee gcc flag from /ghc/compiler/main/DriverFlags.hs
> (gcc-i386 doesn't understand it, and even on Alpha, removing it
> probably wouldn't break the build process itself)

Ok, I assume you'll discover in due course whether you need this on
Alpha or not.

> Commented out:
> 
> #if alpha_TARGET_ARCH
> checkFEDArgs arg_tys
>   = check (integral_args <= 32) err
>   where
>     integral_args = sum [ machRepByteWidth rep
>                         | (rep,hint) <- map typeMachRepRep arg_tys,
>                           hint /= FloatHint ]
>     err = ptext SLIT("On Alpha, I can only handle 4 non-floating-point
> arguments
>  to foreign export dynamic")
> 
> from TcForeign.lhs. typeMachRepRep was nowhere to be found (not in
> ghc-6.4, or any earlier released version), and I wasn't able to
> figure out what, exactly, it does.  It seemed like an RTS thing, so I
> decided to comment it out, and the build proceded.  Not sure if that
> was such a wise idea now. 

That looks like a sanity check that has bitrotted.  You can ignore it
until you need to get 'foreign import "wrapper"' working.

> Commented out:
> 
> import PrimRep		( getPrimRepSize, isFloatingRep )
> 
> from the same file.  The file isn't included with ghc
> anymore, and the imported functions weren't used.

More bitrot.  Please send us patches when you've got the compiler
bootstrapped.

Cheers,
	SImon