On 11/02/2009 08:03 PM, Wayne Meissner wrote:

That does look a bit suspicious.

Given you're packaging for fedora, and fedora has a recent version of
libffi installed (or can be added as a dependency), you could try
creating a patch using changeset
8cbc3f54a875a61e1fb67ac88a25b71940a3ed91

That allows ruby-ffi to build against the system libffi.  Technically
it always allowed this, but ruby-ffi was a bit overly picky about what
it wanted from libffi.

You may find you don't even need that patch.  Just having the proper
libffi-dev packages installed/as dependencies might allow extconf.rb
to pick it up.

You might also need 1d50e0647ac55e15222f9a6a9d163d7a68022186 to
properly detect&  pass powerpc64 into libtest/GNUmakefile, so it
builds for the correct arch.

I  have applied both, and I have gotten past the earlier issue (Thank you!). I am now seeing different errors on ppc and ppc64. On the ppc build [1] I build and get a couple of unit test failures.. On the ppc64 build [2] I see a wierd error about wrong number of arguments.

I apologize, I am not strong in C and did not see anything obvious jump out at me on these.

-- bk


[1] PPC Build Log: http://koji.fedoraproject.org/koji/getfile?taskID=1785189&name=build.log
[2] PPC64 Build Log: http://koji.fedoraproject.org/koji/getfile?taskID=1785192&name=build.log