2.0.104 on sparc

[Philip Blundell <philb@gnu.org>]

Hi,

I built 2.0.104 on my ELC.  Most of `make check' seems OK, though some of the 
floating point tests fail (there seems to be a problem with the underflow 
exception in particular).  But threads seem to have some trouble -- tst-aio.c 
results in one parent process looping forever and a zombie child, and similar 
things happen with nscd.

My knowledge of SPARC assembly is pretty nonexistent so I can't say much more 
about what's wrong.  Perhaps one of the sparc hackers could take a look.

BTW, many of the floating point checks fail on Alpha as well.  Perhaps it 
would be a good idea to have `make check' ignore the results of these tests on 
platforms where failures are expected.  Otherwise the value of "make check" is 
reduced for most users; it would be good if people could use it routinely to 
test a library before installation.

p.