This is the mail archive of the libffi-discuss@sourceware.org mailing list for the libffi project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: libffi - the 10th Anniversary Release


On Jan 29, Anthony Green wrote:
> Eli Barzilay wrote:
> > This is possibly a stupid question -- but aren't the main sources in
> > the gcc subversion repository?
> >   
> It's a good question. What we have here is a case of the tail
> wagging the dog...
> 
> The libffi tree in gcc is definitely the most up to date and well
> maintained tree, however the GCC project views libffi as an external
> project that they include for convenience sake.

Thanks for clarifying.  I was worried because just last week we caught
up with the gcc svn.


> It seems perfectly reasonable to populate the stand-alone libffi
> tree from the GCC project to make a release for users. This simply
> mirrors what some projects using libffi do on their own anyways.

Yes, we do that too -- we have the complete libffi directory (minus
tests), and enough other files (from the gcc toplevel) that make it
possible to run configure.  To do the configure things we alse needed
to drag in some more files, some in the config directory.


> >> In any case, I'd appreciate it if people would like to try checking
> >> out the latest from the libffi cvs repository and build/test it.
> >>     
> >
> > FWIW, we (www.plt-scheme.org) recently updated our copy from the gcc
> > repository, and it builds fine -- we have these patches:
> >
> >  * We recreated libffi/configure with autoconf version 2.61
> >
> >  * We changed config/multi.m4 (not inthe libffi directory) to allow
> >    in-source-tree builds, the change is from line:
> >
> >      if test "$with_target_subdir" != "."; then
> >
> >    to
> >
> >      if test ! -z "$with_target_subdir" && test "$with_target_subdir" != "."; then
> >
> >  * Removed +live_support in src/powerpc/darwin_closure.S so it builds
> >    with Mac OS X 10.
> Thanks for the input. Why do you have to remove the live_support
> section attribute? Is it just because Max OS X 10 tools don't
> recognize it? Is this something we should have a configure test for?

Looks like it works on 10.4, but not on 10.3.  A relevant page that we
found is

  http://www.nabble.com/patch:-gst-2.95h-on-Mac-OS-X-10.3.9-td14569004.html

It would be nice to have a test for 10.4 and above, but IIUC, this is
not a big problem, so we just removed it...

-- 
          ((lambda (x) (x x)) (lambda (x) (x x)))          Eli Barzilay:
                  http://www.barzilay.org/                 Maze is Life!


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]