Re: gmp-3.1.1 make error on MacOS X

[Jon Becker]

Ronald Bruck wrote:

> Koen Vervloesem wrote:
>
> > Thanks for the help, Kevin, but now make still gives errors. I started
> > configure, removed the line you mentioned in config.m4, and then did
> > "make". I have include the output of make as an attachment. Is the
> > "Nothing to be done for 'all'" statement on the 4th line normal? He =
> also
> > complains for a junk character in tmp-add_n.s, and then comes a list =
> of
> > four errors. Are these errors caused by another reason than assembler
> > syntax?
>
> I get the same result; after applying the patch and running 
> ./configure=20=
>
> I get:
>
> [snip]
>
> tmp-add_n.s:64:Junk character -1 (=B7).
> make[3]: *** [add_n.lo] Error 1
> make[2]: *** [all-recursive] Error 1
> make[1]: *** [all-recursive] Error 1
> make: *** [all-recursive-am] Error 2
>
>
> Anybody know what this means, or how to fix it?
>
> I have translated Brian Borchers' SDP solver (csdp) to use gmp, and I'd
> like to get this to working on Darwin.  But first I have to get gmp to
> compile :-)
>
> ---Ron Bruck

In looking through the archives, it appears this was discussed a bit 
back in December.  The problem seems to be that m4 on OS X is broken in 
some way (or perhaps just "different").  The m4wrap_prepend invocation 
causes a 0xFF byte to be written at the end of the tmp-add_n.s file.

I "fixed" this by just removing the invocation of m4wrap_prepend from 
asm-defs.m4.  I'm not sure whether or not that might cause other 
problems, but it did allow gmp to finish compiling.  Apparently it also 
works to install GNU m4, which is no doubt a better solution.  
Ultimately, I would hope that a future release would work around this 
problem with a different definition for m4wrap_prepend or some such.

-Jon