buildbot for buildout (again)
Hello, After the 1.5.0 I tried again to setup a buildbot (and on top of that on *windows*) for buildout. Good news is that the tests now work. Bad news is that there are a hell lot of failures. More bad news is that there are some failures even on linux (a local ubuntu). Does anyone care about having a buildbot for buildout and fixing those failures? -- Best regards, Adam GROSZER mailto:agroszer@gmail.com -- Quote of the day: Kiss your keyboard goodbye!
On Fri, Aug 27, 2010 at 5:13 AM, Adam GROSZER <agroszer@gmail.com> wrote:
Hello,
After the 1.5.0 I tried again to setup a buildbot (and on top of that on *windows*) for buildout. Good news is that the tests now work. Bad news is that there are a hell lot of failures. More bad news is that there are some failures even on linux (a local ubuntu).
Does anyone care about having a buildbot for buildout and fixing those failures?
Obviously, I want the tests to pass. I'll wade in, if necessary after the dust settles from 1.5. I suspect that part of the problem may be that the tests tend to depend on how the python used to run them is configured. I always ran them with a clean Python. Other people tended to run them with a Python that had setuptools or distribute installed, and changed the tests to make them pass in their environment -- and so on. I'm going to take a fresh look at the way the tests are written and run to try to address this, among other things. Jim -- Jim Fulton
Hello Jim, Friday, August 27, 2010, 12:54:12 PM, you wrote: JF> On Fri, Aug 27, 2010 at 5:13 AM, Adam GROSZER <agroszer@gmail.com> wrote:
Hello,
After the 1.5.0 I tried again to setup a buildbot (and on top of that on *windows*) for buildout. Good news is that the tests now work. Bad news is that there are a hell lot of failures. More bad news is that there are some failures even on linux (a local ubuntu).
Does anyone care about having a buildbot for buildout and fixing those failures?
JF> Obviously, I want the tests to pass. I'll wade in, if necessary after the JF> dust settles from 1.5. JF> I suspect that part of the problem may be that the tests tend to JF> depend on how the python used to run them is configured. I always ran JF> them with a clean Python. Other people tended to run them with a Python JF> that had setuptools or distribute installed, and changed the tests to make them JF> pass in their environment -- and so on. I'm going to take a fresh JF> look at the way JF> the tests are written and run to try to address this, among other things. I think it would help a lot having the exact requirements for tests written down. Or even maybe checking those in the tests's setUp? There are some presumptions how it would work, but it's rather hard to guess. -- Best regards, Adam GROSZER mailto:agroszer@gmail.com -- Quote of the day: What happens when you cut back the jungle? It recedes.
On Fri, Aug 27, 2010 at 01:05:29PM +0200, Adam GROSZER wrote:
Hello Jim,
Friday, August 27, 2010, 12:54:12 PM, you wrote: JF> I suspect that part of the problem may be that the tests tend to JF> depend on how the python used to run them is configured. I always ran JF> them with a clean Python. Other people tended to run them with a Python JF> that had setuptools or distribute installed, and changed the tests to make them JF> pass in their environment -- and so on. I'm going to take a fresh JF> look at the way JF> the tests are written and run to try to address this, among other things.
I think it would help a lot having the exact requirements for tests written down.
I believe they are: http://zope3.pov.lt/trac/browser/zc.buildout/trunk/DEVELOPERS.txt
Or even maybe checking those in the tests's setUp?
That would be even better, because people generally don't read documentation. Best of all would be if the tests could be made to work with a Python that already has setuptools/distribute installed. Virtualenv is so much easier than compiling your own clean Python from scratch. Marius Gedminas -- Remember, drive defensively! And of course, the best defense is a good offense!
On Aug 27, 2010, at 6:54 AM, Jim Fulton wrote:
On Fri, Aug 27, 2010 at 5:13 AM, Adam GROSZER <agroszer@gmail.com> wrote:
Hello,
After the 1.5.0 I tried again to setup a buildbot (and on top of that on *windows*) for buildout. Good news is that the tests now work. Bad news is that there are a hell lot of failures. More bad news is that there are some failures even on linux (a local ubuntu).
FWIW, tests passed for me on Ubuntu (Lucid) and Windows when I made a release, with Distribute and Setuptools, with a variety of Python versions.
Does anyone care about having a buildbot for buildout and fixing those failures?
Obviously, I want the tests to pass. I'll wade in, if necessary after the dust settles from 1.5.
I suspect that part of the problem may be that the tests tend to depend on how the python used to run them is configured. I always ran them with a clean Python.
Definitely. I contemplated trying to make it possible to develop buildout with a system Python, but that needs zc.recipe.testrunner upgraded to the new approach of generating scripts, maybe among other changes. And the initial setup necessary is to call dev.py, if that makes a difference. And there are some pre-existing tests that fail intermittently on Linux. I didn't address these. Gary
Other people tended to run them with a Python that had setuptools or distribute installed, and changed the tests to make them pass in their environment -- and so on. I'm going to take a fresh look at the way the tests are written and run to try to address this, among other things.
Jim
-- Jim Fulton _______________________________________________ Distutils-SIG maillist - Distutils-SIG@python.org http://mail.python.org/mailman/listinfo/distutils-sig
Hello, Yeah! Having a clean system python helps a lot. zc_buildout_dev py_254_win32 zc_buildout_dev py_265_win32 zc_buildout_dev py_270_win32 are all GREEN zc_buildout_dev py_244_win32 zc_buildout_dev py_265_win64 zc_buildout_dev py_270_win64 have 1-1 failures. I'll let the winbot now send mails about status to zope-test@zope.org Friday, August 27, 2010, 2:56:20 PM, you wrote: GP> On Aug 27, 2010, at 6:54 AM, Jim Fulton wrote:
On Fri, Aug 27, 2010 at 5:13 AM, Adam GROSZER <agroszer@gmail.com> wrote:
Hello,
After the 1.5.0 I tried again to setup a buildbot (and on top of that on *windows*) for buildout. Good news is that the tests now work. Bad news is that there are a hell lot of failures. More bad news is that there are some failures even on linux (a local ubuntu).
GP> FWIW, tests passed for me on Ubuntu (Lucid) and Windows when I GP> made a release, with Distribute and Setuptools, with a variety of Python versions.
Does anyone care about having a buildbot for buildout and fixing those failures?
Obviously, I want the tests to pass. I'll wade in, if necessary after the dust settles from 1.5.
I suspect that part of the problem may be that the tests tend to depend on how the python used to run them is configured. I always ran them with a clean Python.
GP> Definitely. I contemplated trying to make it possible to develop GP> buildout with a system Python, but that needs zc.recipe.testrunner GP> upgraded to the new approach of generating scripts, maybe among other changes. GP> And the initial setup necessary is to call dev.py, if that makes a difference. GP> And there are some pre-existing tests that fail intermittently on Linux. I didn't address these. GP> Gary
Other people tended to run them with a Python that had setuptools or distribute installed, and changed the tests to make them pass in their environment -- and so on. I'm going to take a fresh look at the way the tests are written and run to try to address this, among other things.
Jim
-- Jim Fulton _______________________________________________ Distutils-SIG maillist - Distutils-SIG@python.org http://mail.python.org/mailman/listinfo/distutils-sig
GP> _______________________________________________ GP> Distutils-SIG maillist - Distutils-SIG@python.org GP> http://mail.python.org/mailman/listinfo/distutils-sig -- Best regards, Adam GROSZER mailto:agroszer@gmail.com -- Quote of the day: BOO! We changed Coke again! BLEAH! BLEAH!
participants (4)
-
Adam GROSZER -
Gary Poster -
Jim Fulton -
Marius Gedminas