Re[2]: Patches to 19981213 config.guess etc for HPPA2.0 supp

[rodneybrown at pmsc dot com]

Sorry, I wasn't clear. The unpatched code failed to recognize our K380  (uname
-m => '9000/800') at all, so failed in config.guess - ending up an unrecognized
triplet of
'-hp-hpux10.20'. Using sysconf to get the architecture is fine, it's just the
pattern
recognizing the machine from uname -m needed to be widened. The model number
matrix was to justify that pattern widening.

egcs-testsuite should have the results for hppa2.0-hp-hpux10.20 after applying
the patches.

Note that config.sub doesn't support hppa2.0{,n,w} if supplied as
a single architecture argument - I only changed the triplet switch statements.

____________________Reply Separator____________________
Subject:    Re: Patches to 19981213 config.guess etc for HPPA2.0 support
Author: <law@cygnus.com >
Date:       17/12/98 22:17


  In message <9812179139.AA913941102@cc.pmsc.com>you write:
  > Heres the evidence for the general 9000/[678][0-9][0-9] pattern
  > for HP numeric model numbers.
I spent some time thinking about the model # stuff, and after looking at the
HP recommended code to get the system's architecture I decided to rely on
the C code to get the architecture.  The fallback code based on model #s is
mostly obsolete now -- I don't really care if the fallback code  mis-identifies
a particular model -- so long as it does so in a safe manner.

  > Also shouldn't {,{gcc,texinfo}/}config.sub be the one hard-linked file?
  > The differences seem to be in stages of evolution, rather than active.
gcc/config.sub has been targeted for termination.  texinfo/config.sub will
probably hang around much longer since we do not control the texinfo package.

jeff