On Tue, 2016-02-23 at 20:06 -0800, Justin Ross wrote: > Hi, everyone. I've been preparing a proof of concept for the reorg, > and > it's reached a stage where it's ready for comment. > > https://github.com/ssorj/qpid-svn-reorg >
Overall looks good. > ## How to setup the Qpid Python dependency > > cd qpid/python/trunk > ./setup.py install --user > export PYTHONPATH=$HOME/.local/lib/python2.7/site-packages > export PATH=$HOME/.local/bin:$PATH We should also fix CMake to find the dependency consistently with the way we find other things: same $PREFIX, pkgconfig, default OS places etc. > It was actually worse than this suggests. Lots of the tests silently > pass (!) if the python testing tools are not available. Some tests are designed to be skipped if optional dependencies are not available. Sounds like we need to strip out some of those checks for things that have become external dependencies but are not optional. > - Gah! Circular dependency. Qpid Python (via qpid.testlib) depends > on qpidtoollibs. This should be removed: Qpid C++ depends on Qpid > Python and not the reverse in any way. Issue here is the blurred line between client and broker tests. We had a load of python tests that ran against qpidd, didn't distinguish testing the client vs. testing the broker, and sometimes assume qpidd specifics (0-10, qmf etc) I would move all that stuff to the cpp broker. That means we need new independent tests for the now separate python component. Thanks to proton we can write client/server tests without an external broker dependency but we need to write them and get the coverage up. > - The new test regime more uniformly runs brokers with valgrind, and > this exposes new valgrind errors. Not surprised. We should fix them. You can write a suppression file as a temporary measure while you reorganize. --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@qpid.apache.org For additional commands, e-mail: users-h...@qpid.apache.org
