> It seems to me that the Lucene tests are about equally stable looking at 
> hudson histories :)

Oh, I didn't mean to differentiate -- I want them BOTH to be running
smoothly. It's just that I can help less with Solr because I know less
about Solr -- that was all I meant to say. Although I did nail that
Long parsing exception recently, if you look at the commit history :)

> I plan on looking into this because it only started on hudson a little while 
> ago - I think perhaps after Yonik did some refactoring around recovery. 
> Clearly something around this recently changed though.

That TestStressNRT in Lucene is fairly new too, I didn't see it
before. Nonetheless, it's getting better.

> But as for all the random occasional fails, thats a much bigger thing than 
> I'm signing up for at the moment :) Hopefully people can chip in where they 
> can as they can on those. The way the tests are these days, there are just a 
> lot of fail paths depending on a lot of variables in both projects. We have 
> worked towards stabilizing them in both projects for years now :)

I do realize how painful it is to debug these things. That's why I
insisted we should try to have tests isolated in the first place to
rule out interdependencies between suites. Solr tests currently leak
out a lot of threads but I haven't looked into how to solve it yet
(but I will). I surely will need help and I will ask for it when the
time comes.

The work on isolation is progressing on my branch here -- LUCENE-3808.
It's been pretty stable (build errors pretty much overlap between
trunk and that branch) -- see the attached screenshot.

Dawid

<<attachment: builds.png>>

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to