From: Carlo M. A. B. <ca...@sa...> - 2008-07-30 06:33:13
|
On Tue, Jul 29, 2008 at 10:36:39PM +0100, Dan...@ba... wrote: > > Looking more closely, I had run configure with --disable-python for > another build before running `make dist-gzip', I suspect that has > something to do with it. I will continue investigating... that is an old BUG that I thought we had make already to difficult to trigger to hurt us, as --disable-python is almost never needed anyway as python support will softly disable itself in most of the cases and the instructions to build a distribution package say to call "./configure" without parameters. Committed revision 1610 for trunk. which should avoid unintentionally removing from the release package files depending on the build environment used for bootstrapping. sadly the fact that we are distributing compiled python bytecodes against all packaging recommendations and building it at bootstrapping time will make the packages different depending if python is available or not, but since most distributions will remove the pyc files and regenerate them anyway that shouldn't be that much of an issue. > Maybe the spec file needs to detect the `disable-python' switch, that is another layering violation, the SPEC should expect a stable package which contains everything needed to build independently of what flags will be used to do so. the current fix should hopefully allow the SPEC to work by luck, but haven't test it. > or maybe the spec file needs to generate two versions of gmond, one with > python and one without (e.g. a ganglia-gmond-nopython package) a --without-python rpm option might be an interesting option for this and will be most likely a good idea for 3.2 which contains 2 versions of gmetad (one of them in python) as well as for a future 3.1 SPEC that will finally consolidate the obsoleted AIX SPEC but definitely not an issue now and not a showstopper for 3.1.0 Carlo |