launchpad-dev team mailing list archive

[Jeroen Vermeulen <jtv@xxxxxxxxxxxxx>]

On 2011-09-30 21:22, curtis Hovey wrote:

> That is indeed what I see and it is repeatable for a set of tests, if
> you change the set, the failing tests are different. It seem that even
> tests that use the librarian (2, 4, ..) always fail.

Also interesting is that in the course of a test run, the blocks of 
nulls get longer every time they appear.

There's been a change to canonical.librarian.storage, but a very rough 
undoing of that change plus killing all the running librarians and 
removing their /tmp dirs doesn't fix the problem.

Speaking of which: any chance that the leaking librarians are part of this?


Jeroen