Re: glibc-2.0.93 homebrew distribution tests

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

>>>>> John Kennedy writes:

John>   Remember, I compiled it on the x686 and copied it onto the x586.  It
John> was never recompiled for the x586.  That was one of my first guesses
John> (which is why I mentioned it) since I see a lot of inline assembly at
John> times for optimized string functions, but since it was time-expensive to
John> test I hadn't done it yet.  I didn't have any problems doing the exact
John> same thing with 2.0.5 thru 2.0.7pre3 so I've been figuring that most of
John> the x686 optimizations didn't affect libc.

John>   Note that I'm using gcc-2.8.1 and binutils-2.9.1.0.4, so I'm not
John> using one of the really blatant optimizing compilers.

John>   If 2.0.93 is doing something new, it is a good time.  They just
John> released linux-2.1.103 so I'll be recompiling everything again anyway
John> (part of my kernel and software endurance test; if it survives compiling
John> everything (3+ hours), it is probably good).  (:

Could you try to configure glibc with an argument of i586-linux to
force it to generate code for i586 (i586 code will run on i686 but not 
the other way round)?

glibc 2.0.93 contains a number of processor specific optimisations
that are not in glibc 2.0.6.  AFAIK there's no special treatment (or
at least very little compared to 2.0.93) in 2.0.7preX for i686.

Did you try to run the programs on i686?  Do they work - and fail only 
on i586?

Andreas
-- 
 Andreas Jaeger   aj@arthur.rhein-neckar.de    jaeger@informatik.uni-kl.de
  for pgp-key finger ajaeger@alma.student.uni-kl.de