PR about glibc 2.1 with --enable-omitfp on i686

[Andreas Jaeger <aj@arthur.rhein-neckar.de>]

PR 956 reports a problem with i686 and -D__USE_STRING_INLINES which is 
enabled by default when using --enable-omitfp.  The problem occurs
with egcs 1.1.1

Has anybody successfully compiled glibc 2.1 on i686 with
--enable-omitfp and egcs?  I just compiled on i486 with
--enable-omitfp using egcs 1.1.1 and had no problems which leads me to 
ask especially for i686.  i686 uses sometimes different routines,
e.g. string inlines, than i486.

Andreas

P.S. here are more details from the report:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
        Hello,

        First of all, I am not using the normal glibcbug report form
because I am not sure if the bug is in your GNU Libc 2.1 or it is an
egcs' bug. The thing is that the library gets miscompiled when using
-D__USE_STRING_INLINES, that is when configuring it with --enable-omitfp
(even changing the -O99 to -O2 or suppresing the -fomit-frame-pointer).
        
        I have tested it with egcs-1.1.1 and egcs-19990208 (2.93.06), on a
Dual PII 450 with 512Mb running Linux 2.2.1-ac5 and binutils-2.9.1.0.19a,
the results I got in each case were the following:

        - when using egcs-1.1.1 the compilation stops with a segmentation 
          fault when trying to run rpcgen on xbootparam.T 

        - when using egcs-19990208 the compilation finishes succesfully, but
          the library is completely broken, every program breaks with a 
          segfault, even those compiled against glibc 2.1 (compiled without
          -D__USE_STRING_INLINES).

        but when I compiled it without the -D__USE_STRING_INLINES it works,
although it breaks some glibc-2.0 programs, for example StarOffice 5.0.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-- 
 Andreas Jaeger   aj@arthur.rhein-neckar.de    jaeger@informatik.uni-kl.de
  for pgp-key finger ajaeger@aixd1.rhrk.uni-kl.de