I've been taking a compilers course, so I think I might be able to help with this project after the school year is over. Seems interesting.
Eric Arc Riley wrote: > So for those who haven't heard, the multicore segfault was finally > traced to Pyrex introducing redundant/unnecessary INCREF/DECREF pairs > on attributes passed to class methods. Cython has fixed this, > apparently, but we can't use Cython due to their project direction > going contrary to the direction we need. They're writing it in the > direction/style for people migrating pure-Python packages to > Pyrex/Cython to make certain functions faster, rather than from the > perspective of "we'd write this in pure C using the Python/C API, but > this makes much more sense". > > PyMill is the solution - and we're getting pretty close to having it > working. Currently mill.distutils imports Cython.Compiler directly > and uses it with a bit of hackery, and it's finding the pxd's alright. > > Now we're down to what changes are needed to get Cython.Compiler > building multiple sources per extension. The current error is "foo > functions declared but not defined", as it expects a 1:1 ratio between > .pxd's and source files. Removing this error will only partly solve > the problem, since the C declarations are in the generated .c files > and not in an extension-wide header. > > I really need more help with this. It doesn't involve any of PySoy's > codebase and is 100% Python code. Yes, it's poorly documented, but > between Piet and what I've learned by digging through the code over > the last two weeks.. > > Join me in #PyMill on Freenode and lets get PySoy stable again! > _______________________________________________ > PySoy-Dev mailing list > [email protected] > http://www.pysoy.org/mailman/listinfo/pysoy-dev > _______________________________________________ PySoy-Dev mailing list [email protected] http://www.pysoy.org/mailman/listinfo/pysoy-dev
