On Thu, Nov 25, 2010 at 19:10, Robert Niederreiter <rnix@squarewave.at> wrote:

Hi Folks,

Why does recent zc.recipe.cmmi sticks hardcoded to version < 1.5.0b1 of zc.buildout in setup.py?

Digging into the logs, I've found that the change is introduced in this changeset: http://svn.zope.org/zc.recipe.cmmi/trunk/?view=rev&rev=118326 (the exact change to that file is here: http://svn.zope.org/zc.recipe.cmmi/trunk/setup.py?rev=118326&r1=115593&r2=11... ). Buildout 1.5 has major differences from previous versions (one being the use of a custom site.py and running python with "-S" option), and that makes it different to test against multiple versions. Currently all the tests for zc.recipe.cmmi pass with Python 2.[4-6] (and zc.buildout 1.4.4), but upgrading to buildout 1.5 introduces three test failures in all three Python versions, due to the changed behavior of buildout. With the current Python trunk (2.7.1rc1+), things are a bit different. Using zc.buildout 1.4.4, the test for zc.recipe.cmmi never finishes, instead it fires up dozens of Python processes in a seemingly infinite loop, and consumes all system resources. Upgrading to buildout 1.5.2, things are a bit better: now the test finishes, but with 4 failures. It seems it is quite hard to write tests against both buildout 1.4.* and 1.5.*, but other than the test failures, the package seems to work just fine with the newest zc.buildout version. Attila