Re: Another fixincludes problem

[Bruce Korb <bkorb at cruzio dot com>]

Richard Earnshaw wrote:
> 
> Bruce, can you advise on this one?  It's been on my to-do list for a
> while, but I've had a local hack to work around it, so I've been putting
> it off.
> 
> We need to start running fixincludes on NetBSD, since we need to fix up
> one or two header files.  One particular case is machine/ansi.h, where we
> need the __builtin_va_list fix.  Unfortunately, this fix is not being
> applied by default.  I suspect that this is because machine is a symlink
> on NetBSD to arm32, the architecture specific directory (this seems to be
> the reverse of freeBSD, where machine is the directory and i386 is the
> symlink).  I really don't want to start adding each netbsd target
> architecture to this fix (it would make the list very long, and constantly
> out of date).  But I have confirmed that adding arm32/ansi.h to this fix
> is sufficient to solve the problem.
> 
> Am I correct in suspecting this is a symlink problem?  If so, have you any
> idea how it might be solved?

Yes, it is.  The wrapper script detects a ``DONE'' file when it
gets around to processing the sym link and skips the tree because
it has already been processed.  The fixincl program only ever sees
``arm32/ansi.h'' get processed.  Add it to the files list for the
machine_ansi_h_va_list hack:

  files = arm32/ansi.h;

Someday in my copious spare time, I'll make the file names
globbing patterns.  But not today.  :-)