Nick Coghlan added the comment: (added Graham Dumpleton to the nosy list to ask if this change may impact mod_wsgi for 3.7)
+1 on the general idea, but given that the current field is a public part of the interpreter state, the replacement access API should really be public as well - we can't be sure folks will always be going through the "PyImport_GetModuleDict()" API. If you replace the addition of the `_PyImport_GetModuleDict` API with a public `PyInterpreterState_GetModuleDict` API, I think that will cover it - the new calls would just be "PyInterpreterState_GetModuleCache(tstate->interp)" rather than `_PyImport_GetModuleDict(tstate)` Folks accessing this field directly can then define their own shim function if PyInterpreterState_GetModuleCache isn't defined. (The rationale for the GetModuleDict -> GetModuleCache change is that "ModuleDict" is ambiguous - every module has a dict. For PyImport_* we're stuck with it, but the "PyImport" prefix at least gives a hint that the reference might be to the sys.modules cache. That affordance doesn't exist for the "PyInterpeterState" prefix. ---------- nosy: +grahamd _______________________________________ Python tracker <rep...@bugs.python.org> <http://bugs.python.org/issue28411> _______________________________________ _______________________________________________ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
