Re: libgmp: link to the correct shared library

[Paul Eggert]

On 7/23/20 5:00 PM, Bruno Haible wrote:

> GNU poke is likely to use GMP soon. When it does that, it will of course use
> the module 'libgmp'. And this package happens to use libtool.

That's fine, all I'm asking for is the ability to use libgmp without using 
havelib. A package that uses havelib shouldn't be affected.

> You can. For uses of system-wide installed GMP, nothing changes. For users who
> built GMP on their own, there is the configure option --with-libgmp-prefix. If
> there is a problem with it, you can redirect the reports to bug-gnulib or me.

That --with-libgmp-prefix worries me, for Emacs in particular. The Emacs folks 
are already somewhat resistent to Gnulib and want to minimize the attendant 
configury. I doubt whether they'd like to add instructions for 
--with-libgmp-prefix, --with-libxyz-prefix, etc. to their documentation. 
Instead, they use pkg-config for many libraries, and this works reasonably well 
for them. I'd rather not add yet another way to do libraries.

I can envision another problem with the havelib approach. For a program like 
Emacs that links many libraries, there is a performance overhead to having a 
long -rpath, so it makes sense to have a single directory (/home/eggert/lib, 
say) for all my personal libraries. But if I do that, then GNU coreutils 
'factor' will have /home/eggert/lib in its rpath whereas 'sort' won't (because 
'sort' doesn't use GMP). And then if I improve glibc and place an improved 
version into /home/eggert/lib, this will affect 'factor' but not 'sort' because 
'factor' is linked with $(LIBGMP) but 'sort' is not. This will be confusing.

There is an irreducible amount of complexity in dynamic linking. I don't much 
like the traditional approach of using -Wl,-rpath in LDFLAGS, and I don't much 
like pkg-config either. It is a pain no matter what we do.

To work around the immediate problem (since this is blocking gnulib integration 
into Emacs) I installed the attached patches into Gnulib. (The first one should 
be routine.) Perhaps we can think of something better.

0001-libgmp-remove-HAVE_GMP-LIB_GMP.patch
Description: Text Data

0002-libgmp-remove-dependency-on-havelib.patch
Description: Text Data