My fellow Qpideers, the state of the CMake union is, uh, well, sort of ok...
A couple of weeks ago I raised the point that enthusiastic commitments to the Cmake transition for the C++ builds early on have dwindled considerably. The Cmake build seems a bit orphaned in Linux land. At the start of the Cmake effort, Andrew Stitcher stepped up and make a bunch of nice improvements to the Cmake files. As time went on, things dwindled somewhat. For the recent past, changes to the Makefile.am files get the corresponding Cmake changes inconsistently. It was recently asked on the list what needs to be done to get Cmake to be the preferred vehicle for builds. Here's what I believe is needed: - The python preprocessing scheme needs to be transferred to Cmake (this is an example of a significant change made to autoconf, but not made to Cmake). Without this, the Cmake test scripts are looking for the python test in the wrong place. (This issue is in QPID-2127) - Someone with experience in Qpid releasing needs to look at what happens during a release and help work out anything needed to get a good release out based on the Cmake system. Also, any doc changes needed for users. - Get someone to try generating an RPM from the Cmake build and see how that compares to the RPMs built via autoconf. Adjust and improve as needed. - Supply more builds to the Cdash dashboard at http://www.riverace.com/CDash-1.4.2/index.php?project=qpid-cpp - I have a RHEL 5 and a Windows XP build there now. The dashboard is driven by outputs from the Cmake build - it makes it very easy to watch the trends, note where tests are failing, make some estimation of what changes caused the problems and who made them, etc. Getting all the build and regression info in one place goes a long way toward helping the entire development crew keep an eye on things. If having this hosted at riverace.com is a problem, I'm available to help migrate it to qpid.apache.org. - Set a deadline for transition to Cmake and deletion of the autoconf files. I understand the fact that for Linux developers, the path of least resistance is to stick with what's known. Therefore, some hard goal is probably what will be needed to make the transition. The goal at the start of the Cmake effort was to get one common build system for all Qpid C++ platforms. That goal was pretty enthusiastically assented to. We're close to a unified system, but in order to keep this from degenerating closer to a two-scheme build system with Visual Studio projects replaced by Cmake files, we need to make a push to get the rest of the way to a single system. Can we get some volunteers to pick up the items above and help complete this effort? Thanks, -Steve --------------------------------------------------------------------- Apache Qpid - AMQP Messaging Implementation Project: http://qpid.apache.org Use/Interact: mailto:[email protected]
