On Mar 18, 2012, at 8:26 AM, Hanspeter Niederstrasser wrote:
> On 03/17/2012 11:20 PM, Daniel Johnson wrote:
>>
>> On Mar 17, 2012, at 9:38 PM, Hanspeter Niederstrasser wrote:
>>
>>> Daniel,
>>>
>>> sip-py24 (need it as a dep for flyway-py) fails to build on 10.5/i386
>>>
>>> c++ -c -pipe -fPIC -Os -w -DNDEBUG -I. -I/sw/include/python2.4 -o bool.o
>>> bool.cpp
>>> c++ -headerpad_max_install_names -bundle -o sip.so siplib.o apiversions.o
>>> descriptors.o qtlib.o threads.o objmap.o voidptr.o bool.o
>>> Undefined symbols:
>>> "_PySequence_GetItem", referenced from:
>>> _parsePass1 in siplib.o
>>> _parsePass2 in siplib.o
>>> "_PyCObject_Type", referenced from:
>>> _PyCObject_Type$non_lazy_ptr in voidptr.o
>>> "_PyBaseObject_Type", referenced from:
>>> _PyBaseObject_Type$non_lazy_ptr in siplib.o
>>> "_PyObject_GC_UnTrack", referenced from:
>>> _forgetObject in siplib.o
>>>
>>>
>>> sip-py25 works, and the equivalent compiler command is
>>>
>>> c++ -headerpad_max_install_names -bundle -L/sw/lib/python2.5/config -ldl
>>> -lpython2.5 -o sip.so siplib.o apiversions.o descriptors.o qtlib.o
>>> threads.o objmap.o voidptr.o bool.o
>>>
>>> Looks like there's no python2.4-config to provide those flags.
>>>
>>> Hanspeter
>>
>> Why do you need -py24? Can you use a newer one? I have no system capable of
>> building python24 (only 10.5 has it) and so can't really try to come up with
>> a fix. At this point, unless someone can come up with a reasonable fix, I'd
>> rather just remove sip-py24 since it doesn't build. The -py24 has just been
>> carried forward through inertia and this is the first failure I heard of
>> even though it must have been broken for a long time. Also, python24 is
>> pretty ancient so I'm unwilling to put much effort into supporting it,
>> especially since I can't run it myself. :)
>
> flyway-py24 (the only existing variant in the 10.5/6 tree) has sip-py24
> as a dependency and this was being done in the context of the stable
> migration.
>
> I did not try to see if flyway-py could be bumped up to a newer python
> version, but I've now cc'd the flyway-py maintainer to see if it can be
> changed from py24 to py25+.
>
> Assuming that sip-py24 needs to be kept in Fink, a hack would be to
> conditionalize the evaluation of `%p/bin/python%type_raw[python]-config
> --ldflags` inside the LFLAGS_PLUGIN chunk of CompileScript so that !24
> uses what's there, and py24 manually adds the needed -L.
>
> The following is the short list of packages have a listed dependency on
> sip-py24
>
> pyqt4-mac-py24
> pyqt-py24
> qscintilla2-qt4-*-py24
>
> I did not follow the dependency tree further down to see what would
> depend on these.
Those are all mine and make up the pyqt4 suite.
No packages directly dep on sip-py24 except for pyqt (for qt3 not 4) and
nothing uses that anymore and I wouldn't be surprised if it doesn't even build
anymore. I'd suggest getting rid it altogether. 3 packages indirectly dep on
sip-py24 through pyqt4: traits-backend-qt-py, spyder{-mac|-x11}-py, and
flyway-py. Only flyway-py has only a py24 variant; the others have more recent
variants too.
I think we should just kill all py24 variants in those packages since there are
so few. If flyway-py can't work with something newer than py24 then it should
go too. It's very old and the upstream site is long gone.
I'm unwilling to put specialized conditional code in sip-py that I can't even
test just for one ancient package. The pyqt4/sip packages are complicated
enough to maintain as it is. :)
Daniel
------------------------------------------------------------------------------
This SF email is sponsosred by:
Try Windows Azure free for 90 days Click Here
http://p.sf.net/sfu/sfd2d-msazure
_______________________________________________
Fink-devel mailing list
Fink-devel@lists.sourceforge.net
List archive:
http://news.gmane.org/gmane.os.apple.fink.devel
Subscription management:
https://lists.sourceforge.net/lists/listinfo/fink-devel
