Re: has makeinfo gone nuts?

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

  In message <19990224093708.D2294@rjlhome.sco.com>you write:
  > > The second part of that might be due to my not noticing yet another
  > > special magical environment variable being passed during a recursive
  > > make.  I probably can't look at that today, but maybe tomorrow -- so
  > > if you can spot anything obvious, let me know!
  > 
  > I'm guessing that's indeed the case.   From the toplevel objdir, a 'make'
  > will DTRT.   It's only when done from within gcc/ that it uses the wrong
  > one.   Perhaps this patch will inspire someone:
The patch seems reasonable to me.  It's similar to how we handle other
host side programs.

God I'll be glad when we clean this mess up.

  > > >But now, on three different systems, I've watched makeinfo consume over 
  > > >hour of compute time and not finish.   truss shows that it's busily mapp
  > > >and unmapping something.
  > > 
  > > No idea on that one.
  > 
  > Whatever that problem was, it's now gone.  I don't see anything in the
  > ChangeLogs that looks obvious, but it works on all three systems now.
This sounds more like the compiler mis-compiled makeinfo at some point.  In
fact, I'm pretty sure I remember that we had a bug like that during the
loop patches saga.


jeff