[Libreoffice-qa] automated testing of Base

[Terrence Enger]

On Thu, 2011-10-06 at 09:31 +0100, Michael Meeks wrote:
> Hi Drew,
> 
> On Thu, 2011-10-06 at 00:57 -0400, drew wrote:
> > > well, kind of - there is a number of features that do not work with ODBC
> > > - so I'm just curious what the reason for using SQLite and ODBC would be
> > > for a test bed?

One motivation is that ODBC data sources can encapsulate
credentials that I ought not to reveal to the whole world.

> 
> 	So - the ideal test code that we have in the build is small, fairly
> self-contained, builds on all platforms and is fast to execute - so we
> can run it every build cycle. That militates against anything using
> Java, and hsqldb in particular. Sqlite OTOH is my preferred replacement
> for hsqldb, and getting it building and being used in the tree would be
> a great start IMHO.

Yes, anything that is part of every build cycle must be
fairly self-contained.  OTOH, it is central to the purpose
of Base that it reach out to the rest of the world: just
look at module name "connectivity".

> 
> > *smile*...maybe I should rephrase that - what could I do to help
> > generate some of the tests?
> 
> 	Heh - so working with Terry & Bernhard to get some VM's setup for
> openQA would be a great idea I think; hopefully it's scaleable.

VMs let us choose what database systems are installed; and
working outside the build cycle relieves the requirement for
"fast", letting us put more value on "thorough" if we so
choose.

Do we have a feeling for what databases are most used "for
real" with Base?  Or how they are used?  Or what open bugs
are most deserving of attention?  How much real use is
through ODBC as compared to JDBC as compared to built-in
drivers?  In short, what will guide us through the many
things that VMs would let us do?


Heh, heh, heh.  I started this thread with the notion that I
would be risking little more than my own time, hopefully
bothering the experts only occasionally with questions like
"Is this worth looking at?" or "Okay, so what should I read
next?" or "Okay, so what should I do next?".  I have time
available unless a customer calls (I wish!), but as the
topic grows wider I see a danger of absorbing more expert
help than my anticipated contribution is worth.  Guidance
welcome.

Terry.