Re: (Fwd) Re: Absolute paths in BFD

[Ian Lance Taylor <ian at zembu dot com>]

   Date: Sun, 30 Apr 2000 09:53:38 -0400 (EDT)
   From: Eli Zaretskii <eliz@delorie.com>

   >    Still, I'm not sure this is a reason good enough to borrow the macros
   >    from GCC, because GCC is quite a different beast.  For example, it is
   >    traditionally case-sensitive to file names, and treats, e.g., foo.C as
   >    a C++ source, even on DOS/Windows.  (The reason is compatibility to
   >    Unix Makefile's.)
   > 
   > That is not relevant to things like IS_DIR_SEPARATOR.  gcc can just
   > ignore issues of file name case, while still paying attention to
   > issues like directory and path separators.

   But then we cannot follow GCC's lead in issues concerning letter case
   in file names, can we?

   Does that mean that it's okay to leave FILENAME_CMP and the related
   changes as I originally submitted them?

Yes.

I never meant to say that we should do precisely what gcc does.
Clearly the programs have different needs.  I meant that we should use
the same macro names when it makes sense.

   >    > Yes, I think setmode could be a problem.  It might be best to test for
   >    > that separately.
   > 
   >    Testing (in the Autoconf sense) could be non-trivial.  Perhaps using
   >    setmode for DOS and Windows (conditioned on appropriate
   >    system-dependent macros, like __MSDOS__ and _WIN32) would be good
   >    enough?
   > 
   > Why is an autoconf test non-trivial?

   Because it turns out there are systems (I think FreeBSD is one of
   them) which have an incompatible function by the same name.

Are there any systems which define O_BINARY but for which setmode is
not compatible with the DOS version?

I believe there are systems which have O_BINARY but don't have
setmode.  That is my concern.

Ian