On 03/11/2011 05:55 AM, Jiri Denemark wrote:
This partially reverts (and fixes that part in a different way) commit e4384459c93e3e786aa483c7f077d1d22148f689, which replaced all examples or scripts used during build to generate other files.
However, python bindings module is compiled and linked against a specific python discovered or explicitly provided in configure phase.
Yeah, I agree that build-time programs and examples should be flexible, but installed python code should be configured.
Thus libvirt.py which is generated and installed into the system we should use the same python binary for which the module has been built.
The hunk in Makefile.am replaces $(srcdir) with $(PYTHON), which might seem wrong but it is not. generator.py didn't use any of its command line arguments so passing $(srcdir) to it was redundant.
Thanks for that comment, otherwise I was really confused.
@@ -1012,7 +1017,7 @@ def buildWrappers(): classes = open("libvirt.py", "w")
extra = open(os.path.join(srcPref,"libvirt-override.py"), "r") - classes.write("#!/usr/bin/env python\n") + classes.write("#! " + python + " -i\n")
Is the addition of -i intentional?
classes.write("#\n") classes.write("# WARNING WARNING WARNING WARNING\n") classes.write("#\n") diff --git a/python/libvirt-override.py b/python/libvirt-override.py index 909ebe3..d544a0e 100644 --- a/python/libvirt-override.py +++ b/python/libvirt-override.py @@ -2,11 +2,6 @@ # Manually written part of python bindings for libvirt #
-# Specify -i commandline option after python was started -if __name__ == "__main__": - import os - os.environ["PYTHONINSPECT"] = "1"
Ah, so it is - now that we don't have /usr/bin/env stealing argv[0], we can go back to the simpler provision of -i via argv[1] (it's a shame that she-bang parsing is so limited). ACK. -- Eric Blake eblake@redhat.com +1-801-349-2682 Libvirt virtualization library http://libvirt.org