Re: underflows, overflows, ... [on Alphas]

[Jeffrey A Law <law at cygnus dot com>]

  In message <rzq67f2mr3t.fsf@djlvig.dl.ac.uk>you write:
  > >>>>> "Jeff" == Jeffrey A Law <law@cygnus.com> writes:
  > 
  >  Jeff> Yup.  It's actually not that bad.
  > 
  > Robert Lipe concluded otherwise, so I'd put the task aside...
Similarly until last night :-)  The chill runtime is pretty simple,
so I figured it was a good hunk of code to try and multilib.

  >  Jeff> You add a few lines of gunk to configure.in files so that it
  >  Jeff> can find the toplevel directory and to conditionally source
  >  Jeff> config-ml.in from the toplevel directory.
  > 
  >  Jeff> Then you wack on Makefile.in in the almost-obvious ways.  Took
  >  Jeff> about 30 minutes for chill to get all the bits close enough to
  >  Jeff> correct that I can't see obvious problems when I build/install
  >  Jeff> the chill runtime.
  > 
  > Can I have a bit more exact specification?  (That's the basic reason
  > it hasn't been done.)  What exactly does one have to end up with for
  > everything else to be happy?
See libchill/configure.in and search for the two hunks of code which
reference topsrcdir.  You'll need to add basically the same code to
libf2c.

In Makefile.in you need to define a handful of multilib variables:

# Multilib support variables.
MULTISRCTOP =
MULTIBUILDTOP =
MULTIDIRS =
MULTISUBDIR =
MULTIDO = true
MULTICLEAN = true

Then look for references to MULTIDO/MULTICLEAN in the various rules.

libf2c will be a little more complicated since it has subdirs.  But
then again, maybe you just need to make the toplevel libf2c stuff
multilib aware.  Dunno.

jeff