your mail

Marc Glisse marc.glisse at inria.fr
Sat Feb 6 23:45:11 CET 2010


On Sat, 6 Feb 2010, William Bader wrote:

> I did search the archives and google, but I must have missed it.
> The archives seem searchable only by month.
> http://gmplib.org/list-archives/gmp-bugs/

http://www.google.fr/search?source=ig&hl=fr&rlz=1G1GGLQ_FRFR359&q=ld.so.1%3A+t-bswap%3A+fatal%3A+%2Fusr%2Flocal%2Flib%2Flibgcc_s.so.1%3A+wrong+ELF+class%3A+ELFCLASS32&meta=lr%3D&aq=f&oq=

or even better with site:gmplib.org among the query terms. (all the "fr" 
are because I am in France, but it works just as well with google.com)

> Even though it is a gcc issue, is there a reason that it isn't documented in
> one of the README, INSTALL or NEWS files or inhttp://gmplib.org/manual/Notes-for-Particular-Systems.html#Notes-for-Partic
> ular-Systems ?

It doesn't happen when you use a version of gcc currently provided by Sun. 
In some sense it does not happen if you have a sane build environment. Any 
64 bit application you try to compile will have the same issue. The only 
difference with gmp is that if you don't specify whether you want 32 or 64 
bits, it defaults to 64 where others default to 32, so more people hit the 
problem (most of them don't actually want a 64 bit version).

> I eventually managed to build a 32 bit version of gcc-4.4.3 for my Solaris
> 10 sparc system by building gmp and mpfr with ABI=32.

How to chose between 32 and 64 bits is well documented. And the easiest 
when you are only trying to build gcc is to let gcc build gmp and mpfr for 
you, as documented in gcc.

-- 
Marc Glisse


More information about the gmp-bugs mailing list