On Fri, Jun 30, 2006 at 11:26:06PM -0400, Tom Lane wrote:
> Something that would actually hold some intellectual interest is to
> improve the testing infrastructure. The current setup is pretty limited
> as to its ability to deal with varying outputs, and even more limited
> in its ability to test concurrent behavior. Again, see the archives.
I must admit I was kind of surprised this didn't generate much
feedback:
http://archives.postgresql.org/pgsql-hackers/2005-08/msg01073.php
It's a simple tester that drives multiple postgres backends
simultaneously and can test whether various concurrently running
transaction correctly block/abort in various situations. I also used it
to test whether all the different LOCK types work as documented in the
documentation (they do, the program checked all combinations of two
locks).
It would in theory be possible to generate scripts to test thing like
simultaneously firing multiple CREATE INDEX commands in seperate
transactions and see how they interact.
Have a nice day,
--
Martijn van Oosterhout <kleptog@svana.org> http://svana.org/kleptog/
> From each according to his ability. To each according to his ability to litigate.