Re: Including GMP/MPFR in GCC repository?

[Mark Mitchell <mark at codesourcery dot com>]

Richard Earnshaw wrote:

> I think there's a very important distinction that needs to be drawn
> between a tool that needs to be installed to *build* gcc and a tool that
> needs to be installed in order to *run* gcc.  GMP/MPFR is needed for the
> latter; and to date we have never relied on such an external component
> that isn't part of the base installation of a system (such as libc).

True, and I agree that this is a substantial distinction.  However, I 
don't think we should distribute even libraries required to run GCC as 
part of GCC itself.  Of course, I have no objection to putting 
known-good source tarballs for such libraries somewhere readily 
available, etc.  But, I think the GCC distribution should not directly 
contain these.

It's true that if you link the compiler itself against a shared library, 
you have to either set the dynamic search patch, link statically, or 
require users set LD_LIBRARY_PATH -- but you have to do that regardless 
of whether or not the libraries are part of the GCC source distribution.

As to zlib and intl, if I had my way I'd remove them.  However, I 
understand that we have various entanglements (as Daniel mentions) that 
may prevent that.  I certainly don't think removing zlib from our 
repository is the most important improvement we can make to GCC. :-) 
But, I do think we should resist incorporating more external components 
into the GCC repository and into its own build process.

--
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713