Re: Building gmp 3.0

[Kevin Ryde]

Roger Highmore <address@hidden> writes:
>
> 1) I am unable to find .c files which provide
> implementations of the four functions mpn_add,
> mpn_add_1, mpn_sub and mpn_sub_1. Please, where in the
> gmp 3.0 distribution can I find these implementations?

They're compiled by mpn/inlines.c, by asking gmp.h for the code.

> 1) Several files are missing the #if
> __STDC__…#else…#endif constructs which distinguish
> between ANSI C and K&R C function prototypes.

Thanks, we've now eliminated that stuff and use automake ansi2knr to
support old compilers.

> 2) I had to add #include<stdlib.h> to several files in
> order that the compiler would recognize calls to
> abort().

Thanks, I think most of those have been fixed.  Normally harmless in
any case.

> 3) I had to add prototypes of mpn_kara-sqr_n and
> mpn_toom3_sqr_n to mpn_mul.c in order that the
> compiler would recognize these.

Fixed in version 3.1.1 I think.

> 4) I had to rename the variable ‘this’ in
> stack-alloc.c, in order that it would not be confused
> with the reserved meaning in C++.

Yes, I noticed the same thing, fixed in 3.1.1.  Of course it's C code,
not C++, but at one stage we tried compiling as C++ just for kicks.

> 5) I had to remove the prototype of strtol from
> set-str.c, and add #include <stdlib.h>, in order to
> resolve an error where the linker was unable to find
> an implementation of strtol.

Thanks, yes, fixed in the development sources.

> 6) I had to remove ‘const’ from mp_clz_tab.c and
> longlong.h in order to eliminate an error where the
> linker was unable to find an implementation of
> __clz_tab[].

Ah, ok, I think MS C doesn't give a non-zero __STDC__ so it misses out
on "const" in longlong.h.  That's been addressed in the development
sources.  Don't know why a missing "const" would make the linker fail.