On Mon, Oct 15, 2012 at 05:02:48PM +0200, Landry Breuil wrote: > On Mon, Oct 15, 2012 at 10:47:08AM -0400, Stuart Cassoff wrote: > > No interest? > > not really, but i'll try to run it through a bulk build. > you should see the guard_local thingy with matthew@, but maybe the > machine where you tested wasnt enough up to date.
I've also experienced the __guard_local failures on an up2date ad64, some python regress tests fail: /usr/local/bin/python2.7:/usr/obj/ports/swig-2.0.8/swig-2.0.8/Examples/python/constants/_example.so:undefined symbol '__guard_local' The update itself only broke zarafa-webaccess: icalmapi_wrap.cxx:3678: error: 'SWIG_From_long' was not declared in this scope So see that with zarafa maintainers. that'd also be good to have it tested on ppc & sparc64. Landry > > On 10/03/12 06:34, Stuart Cassoff wrote: > > > On 09/08/12 19:58, Juan Francisco Cantero Hurtado wrote: > > >> On Sun, Sep 09, 2012 at 01:01:37AM +0200, Juan Francisco Cantero Hurtado > > >> wrote: > > >>> On Fri, Sep 07, 2012 at 06:28:12PM -0400, Stuart Cassoff wrote: > > >>>> Now that you're in there, why not bring swig up to the latest? > > >>>> > > >>>> I posted a diff for a swig-wip a while back and looking at it now, > > >>>> it doesn't seem too hard to upgrade. I do remember being confused > > >>>> about boost and that maybe swig will find and use things if > > >>>> installed, even if --without-things is used. > > >>> > > >>> I guessed that exists a good reason for the outdated port, so I only > > >>> added the line with the option. > > >>> > > >>> I've been reading the changes of the last four years and working in > > >>> the update this afternoon. The update isn't so easy, 68 packages > > >>> depends of swig and I've seen at least 4 patches related to swig. > > >>> > > >>> Also I'm explicitely disabling the most of the languages (except the > > >>> enabled in 1.3.6). I don't like the magic of "configure" for this > > >>> package. If someone wants enable some language, he or she needs add > > >>> the necessary stuff to the Makefile (modules, *_depends, etc). I may > > >>> be wrong, but I think that dpb will generate different packages with > > >>> each bulk build if we rely on the magic of "configure", eg: if dpb is > > >>> compiling swig and packageY (that depends of languageZ) at the same > > >>> time, swig will enable the support for languageZ because the package > > >>> is installed in the build machine. > > >>> > > >>> I can't do a bulk build for to test the packages that depends of swig, > > >>> but I'll review the makefiles for to see the languages necessary. I'll > > >>> send you the patch the next week :) > > >> > > >> I was a bit wrong. The dependencies are only necessary for the tests and > > >> examples, not for compile the package or to use this from other > > >> languages. > > >> > > >> Don't blame me, I haven't used swig before of today :) > > >> > > > > > > Latest diff, based on your work and mine. > > > Swig now seems to need boost. Objections/Suggestions? > > > Some tests fail because of '__guard_local'. > > > http://www.openbsd.org/cgi-bin/cvsweb/src/gnu/gcc/gcc/targhooks.c rev 1.3 > > > I have no idea what to do about that. > > > Tested on amd64 and with a Tcl extension that wraps OpenGL (a big wrap). > > > Not tested in a bulk or with other extensions. > > > > > > Stu > >
