This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: config.sub pentium2 and pentiumii aliases (fwd)

  > We certainly are providing "a compiler for the Pentium", which
  > sounds quite like "a compiler for the Objective C language".
  > If we do want to avoid trademarks, then it would be better to have
  > completely arbitrary names, rather than bogus ones like 686 as
  > it is used in gcc, which can only confuse.
  > And somewhere the documentation will have to say 686=Pentium-3, there
  > is simply no way of avoiding *some* reference to the architecture for which
  > the compiler is designed!

For gcc, yes. What about all of the other packages that use
config.{guess,sub}? The majority of which don't *care* which x86-family
CPU you use, just that it's i386+. 

As far as '686' being bogus, I have to disagree to a point. It simply
means "family 3, class x86". It doesn't specify the model or stepping
level, granted, but "pentiumiii" doesn't in any
easily-parsable-by-configury form. 

Personally, I'd like to see 'config.guess' spewing i386-*-* (or something 
similar - ix86, ia32, and x86 have also all been suggested), and a set of
autoconf tests added for detection of family, model and stepping levels
for those (relatively few) applications that need them.

In the meantime, Linux tells me that I'm using an i686 - I use
a Coppermine Pentium III - so if pentiumii is i786, I've got an issue ;)

As I recall, config.guess uses what 'uname -m' tells it - on Pentium II
and Pentium Pro systems, this says 'i686'. If config.guess says 'i686',
then config.sub should say the same thing when passed an alias describing
this type of system (unless config.guess or uname are wrong, but that is
a different issue).

Mo McKinlay
GnuPG/PGP Key: pub  1024D/76A275F9 2000-07-22

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]