I agree with all the suggestions. It shouldn't be too much trouble to
implement. I'll do it sometime after I am done with Stored Procedure
support for Beta. Build script is in Perl and therefore can be easily
customized to do whatever ;-).
1. Doing unit tests on all databases: totally doable, since now I have
all those databases on my home machines
2. Uploading reports: don't see any problems, another call to rsync
that is used for uploading tarballs.
3. Performance Tests: this is not all straightforward. I noticed that
the number can change dramatically if I change the order of "suites". I
guess Java class loading and garbage collection are the factors that
add to uncertainty. I guess we need to run each suite in a separate
JVM. I guess I can do some ant based runner. In general I am hesitant
about publishing performance numbers for the development code. This may
create a wrong impression, and lead to premature optimization attempts.
4. Regression. I believe there were some Oracle-specific issues. Don't
remember now.
Andrus
On Saturday, March 22, 2003, at 06:55 AM, Holger Hoffstätte wrote:
>
> It would be nice if the nightly builds would be run for all available
> databases, and also if the tarball directory would contain the reports
> for
> all the tested databases, even if everything has been successfull. This
> would allow everybody to see whether a regression affects all
> databases or
> just a specific adapter, etc. Generating reports even when everything
> works would give everybody a good feeling for free. :)
>
> Would this be possible or too much hassle? How about the performance
> test
> and the regression app - could/should we run these nightly as well?
> Does
> the regression app run only on Oracle or should it work with other
> DBs? (I
> hope so!)
>
> Holger
>
This archive was generated by hypermail 2.0.0 : Sat Mar 22 2003 - 14:49:26 EST