egcs 1.0 and hp-ux 10.20 on a D class.

[Jeffrey A Law]

  In message < m0xkZtt-00047LC@psg.com >you write:
  > I've recently built egcs 1.0 on an HP D class box running 10.20.  I
  > had a few problems, relatively minor I think.
Good to hear from you.

  > First, configure mistakenly (I think) identifies the machine as PA
  > 1.0.  I don't know if this is important, or who I should be reporting
  > it to.  The problem is in config.guess.  Uname identifies the machine
  > as a 9000/803, which the case statement in config.guess thinks is hppa
  > 1.0.
Thanks.  I've fixed this for the next snapshot.


  > The build goes fine.  When I ran make check, I did get one bad
  > response.
  > 
  > >FAIL: gcc.c-torture/execute/ieee/fp-cmp-1.c execution,  -O0 
Known failure for PAs.

  > It also prints out the following repeatedly, which I assume is
  > unimportant:
  > 
  > >WARNING: Couldn't find the global config file.
Just a warning from dejagnu, it's nothing to worry about.

  > Otherwise the tests seemed fine.  Actually, that puzzles me, shouldn't
  > I have gotten some errors with exception handling on hp?  I thought I
  > saw something about that on the mailing list...
If you're using shared libraries or the HP assembler, then you're likely to
have problems.

  > Lastly, I noticed that egcs (and gcc) gives the size of a long double
  > as 8 bytes, while the HP C compiler uses 16 bytes.  Not a big deal,
  > I was just curious why that was so.
Nobody's bothered to implement long doubles or test them on PAs.

jeff