GMP and clang bugginess

[Torbjörn Granlund]

bodrato at mail.dm.unipi.it writes:

  Do we have any working configuration for the x32 ABI?

It works.  We had testing of it on a Gentoo system until perhaps a year
ago.

The reason for the ivydeb32v7.gmplib.org-stat-clang-clang++:x32 failure
is that clang apparently accepts and ignores -mx32.  Autoconf relies on
that things that don't work fail visibly, an accept-and-ignore policy
clashes with autoconf.

I am not worried about that type of failures, they leave GMP users in no
doubt.  I do worry that about 18 of our 94 clang configs silently create
a miscompiled library.

What fraction of users run "make check".  I have no idea about the
figures, but I expect very very few do that.  Fortunately, most users
today probably use a pre-built GMP, and package maintainers are likely
to both use gcc and to run "make check".

Compiler bugs is not a new problem for GMP.  GCC also used to have
problems before GMP was a requirement for GCC.  But it was never even
close to clang in bugginess.

-- 
Torbjörn
Please encrypt, key id 0xC8601622