On 27/06/12 21:59, Daniel Kinzler wrote:
If we decide to do this, there should be some way to
define and re-create the
initial contents of the read-only database. Perhaps a maintenance script that
other parts of the code (and extensions) can register with could do the trick.
What do you think? Is this feasible?
-- daniel
How do you define a database read? What would they read if there had
been nothing written earlier?
The idea seems fine, but if we are not really interested in setting
somnehting to be read (ie. we don't care about the read results), that
suggests that it should be possible to make the test run without a db
(eg. see patchset 13037 for one change of that kind).