Re: fixincludes bug?

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

  In message <387D17A2.8D037D1B@sco.com>you write:
  > Jim Blandy wrote:
  > > 
  > > Below, I've got the log entry and diff for revision 1.17 of
  > > egcs/gcc/fixinc/mkfixinc.sh in the sourceware.cygnus.com EGCS
  > > repository.  Does that log entry really describe that change?
  > 
  > Not sufficiently.  The change enabled the fixinclude processing
  > from being a no-op (disabled) *and* added a fix for the __FD_ZERO bug.
  > The buggy `//' comment handling was fixed in an other patch that
  > addresses the problem in C code.  `sed' is just not up to the task
  > of keeping track of sufficient context.
  > 
  > > Either way, the change to the i[34567]86-*-linux-gnu* case causes the
  > > build process to default to egcs/gcc/fixinc.sh, the old-style fixinc
  > > script with the buggy `//' comment handling.  Is this really what was
  > > intended?
  > 
  > Yes and no  :-).  fixincludes is now supposed to be enabled for IA86 Linux,
  > but it should be using the working `no_double_slash_fix' routine
  > in fixfixes.c.   Is that routine failing?  If so, can you send me
  > a fragment or file example where it fails?  Thanks!
Red Hat is using the generated script.  That's what's failing.   The
generated fixinc script needs to work just as well as the C version.

jeff