other/2620: Filenames differ only in case

[David.Billinghurst@riotinto.com]

>Number:         2620
>Category:       other
>Synopsis:       Filenames differ only in case
>Confidential:   no
>Severity:       non-critical
>Priority:       low
>Responsible:    unassigned
>State:          open
>Class:          change-request
>Submitter-Id:   net
>Arrival-Date:   Mon Apr 23 21:26:00 PDT 2001
>Closed-Date:
>Last-Modified:
>Originator:     David Billinghurst
>Release:        gcc 3.x
>Organization:
>Environment:
all
>Description:
This summarizes discussion on gcc@gcc.gnu.org on/about 11 April 2001.  It is a minor annoyance that should be fixed one day.

In the source tree we have 
*	libffi/src/mips/n32.S
*	libffi/src/mips/n32.s
*	libffi/src/mips/o32.S
*	libffi/src/mips/o32.s

The *.s just include the *.S

This causes some warnings when using CVS on cygwin, but fortunately the files aren't used on cygwin.  There is the potential to cause problems building a cross compiler, etc.

Tom Tromey suggested a fix

>How-To-Repeat:

>Fix:
>From Tom Tromey <tromey@redhat.com>:

My suggestion is to copy *.S to *.s (if you look at the .s files you'll see they are just #include <> anyway), update Makefile.am, and make sure to pass `-x assembler-with-cpp' to the gcc invocation (we know only gcc will be used to build libffi) somehow.  (You might think we could just use *.S but I think that is unreliable.  I've found in practice that sometimes the case of a file can seemingly randomly be
changed on NT at least.)

I would do this but as it turns out I don't immediately have a convenient way to test this.  Maybe eventually I'll build a cross toolchain to try it.  I'd rather someone who builds on the right sort of machine regularly do it though.

Tom
>Release-Note:
>Audit-Trail:
>Unformatted: