Michael Ströder wrote:
Howard Chu wrote:
Hmm, another 100 iterations without a problem. Not sure whether it's related to back-mdb or whether this may happen occasionally with other backends too. Trying more...
Looks like a simple timing dependency in the test script. slapcat of the DB in your testrun directory shows that the correct value is present. The refint overlay does its work in a separate task, so there's no guarantee it will be done by the time the original modrdn request finishes.
We could add a SLEEP in the script here but it seems to be such a rare occurrence I'm inclined to ignore it.
Hmm, I'd argue that a SLEEP would be helpful to avoid further questions about it. A OpenLDAP packager might run "make test" routinely and this issue aborts it.
I've replaced the hardcoded "sleep 1" in the script with "sleep $SLEEP0" so it can be overridden along with the other SLEEP1/SLEEP2 values.
Packagers running the tests should either ensure they're on an idle enough machine, or set larger values in their environment...