Update for swig/python #4093
Update for swig/python #4093
Conversation
this adds the module with the functions to disable and re-enable the #include<pj/compat/socket.h> for win32 systems
adds the transform_docs solution to the setup.py file, correcting the pjsua2.py file for IDE comments
adding the transform function for setup.py to transform pjsua2.py
adding functionality from win32fix to ask the user if the path is correct, if not, type the path where it should be from the projects main folder. it also includes new checkers for win32, and if it is using mingw64 version, or any python path that is currently being used. it expects you have installed pycparser and pycparser_fake_libc via pip. if it cant find any matching, it asks you for the full path to pycparser.
|
|
|
I tested it here on Mac and it didn't work either. Unfortunately, I also don't know how to fix it. |
I have tried multiple things, and yes i have followed the build steps to the letter from the guides "they seem sorta lacking when it comes to pjsua2 swig building, like no mention of c++xx version for compiler, which version of pycparser or python 2.x version to use" which could be the potential issue here? would need more info from those who work on the swig/python side of things to clear it up. when I use make, I do also get similar result as this bug report does: #4092 (comment) so if I really want to build the module, I have to just trust a potentially outdated symbolic link file and just force swig to process it. |
i made changes to make importsym more adaptive, along with checking if the system is win32, if it is, it will ask the user the registered path is correct, if not change it to the correct one, then edit and comment #include<pj/compat/socket.h> for them.
at the same time made changes to setup.py, including automatic execution of transform_docs.py, adding code comments for the user without any hassle.
I have tested the changes against linux and windows, and they seem to function as expected.
however i do still happen to hit a roadblock of the actual process function having issues with _fake_typedefs.h "see screenshot for reference.", which is likely just a setup issue on my end/versioning issues....
Linux version:
Windows version:
there is also probably some methods i could have done better, tho i would like to hear feedback on the solution!