hourly builds on hudson

[jmesnil]

I'v noticed that hudson now build and run JBM 2 tests every hour in addition to run when when the code is modified.

Is it on purpose? I find it noisy and I'd prefer we switch back to running only on code modification.

[timfox]

Noisy in what way? You only get an email if it fails.

It needs to be run repeatedly since there are some race conditions that only show up every X runs, where X can be quite large. This allows us to catch them.

[jmesnil]

"timfox" wrote:
Noisy in what way? You only get an email if it fails.

Noisy when I browse Hudson UI to correlate a failure with a svn changeset and there are builds in-between.

"timfox" wrote:

It needs to be run repeatedly since there are some race conditions that only show up every X runs, where X can be quite large. This allows us to catch them.

ok. if the change was done on purpose, I'm fine with it.

[timfox]

Yeah, the continue test is deliberate.

As I say it helps us catch race conditions and other intermittent errors that might only show up in 100 runs. This has already shown up quite a few intermittent and difficult to catch issues.

You could ask Alex if he'd liked to add a feature to Hudson to show you the last changes without having to click back through previous builds. That would be useful.