On 01/02/2013 11:53 AM, Julien Cristau wrote: > On Wed, Dec 12, 2012 at 00:55:27 +0100, Julian Taylor wrote: > >>> They were generated with 2.0.1 and 2.0.4 vs 2.0.7 in testing >>> the difference is mostly bugfixes and minor enhancements according to >>> swig release notes. >>> >> >> but this does not look so good: >> SWIG-2.0.8 summary: >> - Fix a couple of regressions introduced in 2.0.5 and 2.0.7. >> - Minor fixes/enhancements for C#, Java, Octave, Perl and Python. >> >> swig2.0.8 is only in unstable and there is no unblock filed :/ >> > Have you tried comparing the files generated by swig 2.0.7 vs 2.0.8 for > scipy? > > Thanks, > Julien >
the difference is only catching of memory allocation failures of this type:
#if PY_VERSION_HEX >= 0x03000000
inst = PyBaseObject_Type.tp_new((PyTypeObject*) data->newargs,
Py_None, Py_None);
- PyObject_SetAttr(inst, SWIG_This(), swig_this);
- Py_TYPE(inst)->tp_flags &= ~Py_TPFLAGS_VALID_VERSION_TAG;
+ if (inst) {
+ PyObject_SetAttr(inst, SWIG_This(), swig_this);
+ Py_TYPE(inst)->tp_flags &= ~Py_TPFLAGS_VALID_VERSION_TAG;
+ }
I see no difference to 2.0.4 in these codeparts. So 2.0.7 should be fine
in terms of regressions.
attached the full diff
swig.diff.xz
Description: application/xz
