From: William S F. <ws...@fu...> - 2008-09-12 22:54:22
|
Haoyu Bai wrote: > Hi, > > I have merged my Python 3.0 support branch to SWIG trunk, and make > sure that all the Python test cases passed. And I haven't run tests on > other modules, so I don't sure whether I've accidentally broken > anything. > > I'd say that the merge progress is not very smooth, even under the > help of svnmerge.py. The instruction of using svnmerge.py to merge > back to trunk could be seen at here: > http://www.orcaware.com/svn/wiki/Svnmerge.py#Merging_development_branches_back_to_trunk > > However, svnmerge.py's merge progress is very slow, and finally it > give an incompleted merge and some conflcts for me, that's maybe due > to the slow connection to SourceForge. So I have to resolve conflicts > by hand, and compare the trunk and my branch with diff. And then, I > did the merge which svnmerge haven't done by directly copying the > files from my branch to trunk. > > Well, hope other merges are luckier than mine.:) > Hi Haoyu Thanks for merging over. I think you have to use svnmerge.py from the start, otherwise you get svn 1.4 usual crap merging. I've done a review of your merge to trunk, just when you thought your GSoC work was done, I've a few minor requests ;) I don't have python 3 installed. Is the output from make check-python3-test-suite actually completely warning and error free? It looks beautifully quiet using 2.5. I saw some code to run the examples. Is it possible to run them all at once as a separate step after 'make check-python-examples'? If so we can use this model for all the other languages at some point in the future. I can't see a clean target cleaning up the python files generated by 2to3 in the examples nor test-suite. Can you add please? Please add to the documentation any limitations, eg overloaded methods don't get annotated. This particular limitation is quite a biggy. Any plans to fix? I think there might be a merge problem in Examples/test-suite/python/Makefile.in as USE_VALGRIND was moved to common.mk. You've checked in pybuf_benchmark_runme3.py and some other runme3.py files. I thought all the runme3.py files were generated, but I guess these are python 3 only tests? Some of these files also look like they display to stdout. You can leave code in which writes to stdout by setting a debug variable of some sort in the test, but it shouldn't be the default. The new function PySwigObject_richcompare goes against the Python naming conventions, ie it starts with Py. Actually we have a few of those in SWIG which need renaming. I don't see any documentation in CHANGES.current for the cscanner.c and parser.y changes. Is this a bug fix? Thanks again for a great contribution to SWIG. William |