Hi again Koha developers - Galen and I just submitted a few patches that implement this new testing method. I'm interested in any feedback you have before they're committed. http://lists.koha.org/pipermail/koha-patches/2008-April/000172.html http://lists.koha.org/pipermail/koha-patches/2008-April/000171.html http://lists.koha.org/pipermail/koha-patches/2008-April/000173.html http://lists.koha.org/pipermail/koha-patches/2008-April/000174.html http://lists.koha.org/pipermail/koha-patches/2008-April/000175.html These create a new t/Makefile which runs a t/database_dependent.pl test script which creates a test database and runs some automated tests against it. Those tests are in the new t/lib directory and use Test::Class to organize and run them. The way that you use this is to re-run the top level 'perl Makefile.PL' which will prompt you for some information about a test database that can be built. Then, you can 'cd t' and run 'make test' in there. This will make sure that your new test database is created and used for the test suite. I expect that eventually this process will be made easier, but this is a first stab at a solution, and I'm trying to avoid interfering with normal installation as much as possible. I'm interested in hearing suggestions about how the process of setting up the database and running the zebra stuff works. I'm also interested in hearing about any of these tests that fail. I may take the opportunity to fix some of these before the code is committed, or fix them later. Thanks for your help. -Andy On Thu, Mar 27, 2008 at 1:43 PM, Andrew Moore <andrew.moore@liblime.com> wrote:

Hi Koha developers -

I have caught some of you on IRC and such, but for the rest of you, I'm a new developer at LibLime working on Koha.

I'm interested in improving the automated test suite a little bit, and I'm hoping to get some feedback from any interested developers and users. There are some database specific tests in the t/db_dependent directory, but I think they can be improved quite a bit. I'll lay out what I'm picturing, and please let me know if you picture a different plan, or have any ways to improve this one.

I'd like to have a 'make' target similar to 'make test' that would build a nearly empty, skeleton database that we could run some automated tests against. Since this would not be the same database that live data is held in, we wouldn't have to worry about deleting or overwriting any useful data. The testing could insert example data and data known to trigger past bugs to ensure that they don't recur. I don't imagine making this a standard part of 'make test', at least at first, since I don't imagine that all end users want to have a second database built just to test against. I think it could be immensely useful for us developers, though, right away.

One useful way I've seen this done in the past is trough the use of the Test::Class perl module. It's similar to the xUnit framework (or jUnit or nUnit) in that it lets you run a bit of code before and after each test to set up the environment (the test fixture). That way, you can ensure that the database looks clean between when different pieces of the test suite runs so that one test script doesn't interfere with another. It also lets your test scripts inherit from each other so that similar test scripts can share the same code to set up their fixtures.

I have a small working prototype of this, but before I get much further, I want to ask if you have any suggestions or advice for this. Please let me know if this sounds like something that would be useful to you, and if there's anything I need to make sure I consider.

Thanks, -Andy

_______________________________________________ Koha-devel mailing list Koha-devel@nongnu.org http://lists.nongnu.org/mailman/listinfo/koha-devel