Stefan Behnel added the comment: Hi, thanks for bringing in the 'historical details'. It's not so much that "Cython has been relying on it" - it's entirely up to users what they compile and what not. It's more that I don't see anything being wrong with it as a feature and that it worked before.
The reason why I found it was that I'm trying to make Python benchmark suite run better in Cython, and that requires compiling all code, some of which is often hidden in the package file. I understand that this is probably a rare real-world requirement because things can usually be made to work by moving code into a separate extension module and reimporting it from there into an __init__.py module. I guess that's why people told you to drop the feature. There's also issue13429, which still restricts the init time compatibility of binary extensions with normal Python modules. Arguably, to make __init__.so work correctly, the "__path__" attribute would have to be set at module creation time as well, otherwise, relative imports won't work during initialisation. I guess I should update my patch in that ticket to reflect that. ---------- components: +Library (Lib) -Interpreter Core _______________________________________ Python tracker <rep...@bugs.python.org> <http://bugs.python.org/issue15576> _______________________________________ _______________________________________________ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
