libgcrypt 1.4.6 mpi on amd build fails?

Werner Koch wk at
Wed Jan 19 21:08:06 CET 2011

On Wed, 19 Jan 2011 19:52, simon at said:

> Why is the asm stuff enabled in this situation?  Instead, how about
> whitelisting the few combinations that are known to work (i.e.,
> non-cross compiles on modern and tested systems) and let people building

We do exactly this (mpi/config.links).  However if config.guess guesses
the wrong system we can't do much about it.  We want to fix either
config.guess to detect the system properly or the autoconf cross
compiler test.  This kk1 version is a Debian build of libgcrypt and they
usually replace config.sub with modern versions, but I am not sure about
this specific one.

In any case, a system with a 64 bit kernel and a 32 bit user land is
non-standard (although common it seems).



Die Gedanken sind frei.  Ausnahmen regelt ein Bundesgesetz.

More information about the Gnupg-devel mailing list