On Thursday, August 21, 2003, at 11:50 pm, Michael G Schwern wrote:

On Thu, Aug 21, 2003 at 10:19:35PM +0100, Fergal Daly wrote:
[snip]
Also you can allocate a sub block to each thread and you don't have to worry
about it's output getting confused with the output of any other thread
because every thing from thread 1 will have a number starting with 1. and
thread 2 will have 2. etc

This is convincing.

Another use case would be the distributed execution of a test suite.


A few months ago I had an test suite that took so long to execute (20-30 minutes) it was interfering with integration (I like to run the tests every time somebody checks in).

My solution was to distribute the test execution over several machines. However, I could only do this at the test script level since I needed to feed a plans worth of tests to Test::Harness at a time, which meant that test output blocked all of the time waiting for one script or another to finish.

It would have been nice to have been able to interleave output from several different plans so we could have got to see failing tests more quickly.

Adrian

Reply via email to