Over the past couple of weeks, I've been working on getting the test server at http://ci.tesla.usability.wikimedia.org/ up and running on a regular basis.
To do this, I've had to scale back the static code analysis for now, so the old style checks (which don't yet target the MW style, so may not be too useful), as well as other, more complicated static code analysis, may not be working when you look.
In the mean time, I've put the configuration of the server under version control at
http://svn.wikimedia.org/svnroot/mediawiki/trunk/test-server/
Soon, I'll have the build server bootstrapping from the configuration in SVN so that anyone can help maintain it if they have commit access to that part of the tree.
In the meantime, I've worked with MZMcBride to so that the codurr bot in #mediawiki announces whenever the tests (including the parser tests) break along with the possible culprits in terms of committers/commits.
Since not everyone's IRC nick is the same as their commit ID (for example, “hexmode” is my IRC nick and “mah” is my commit ID), there is an attempt to use the USERINFO files to translate from commit ID to IRC nick. To take advantage of this, update your USERINFO file. See http://svn.wikimedia.org/svnroot/mediawiki/USERINFO/mah for an example.
Finally, as a preview, this week I'm starting to integrate the Selenium tests into the PHPUnit run.
Mark.
On 31.08.2010, 20:58 Mark wrote:
Finally, as a preview, this week I'm starting to integrate the Selenium tests into the PHPUnit run.
How slow are they? We've been struggling to make people run something other than ol' good parser tests before committing, and the slower are tests, the less desire people will have to run them, even if their systems can run Selenium. Therefore, I propose that Selenuim should be disabled by default, but be easily enableable per-user or per-run (with a command-line switch).
Max Semenik maxsem.wiki@gmail.com writes:
On 31.08.2010, 20:58 Mark wrote:
Finally, as a preview, this week I'm starting to integrate the Selenium tests into the PHPUnit run.
How slow are they?
They don't seem especially fast. But given that most people aren't going to have Selenium set up for, the tests would fail anyway as there would be no way to test them.
So I agree, Selenium tests should only be run when the user requests them.
Mark.
wikitech-l@lists.wikimedia.org