I was investigating why I was getting intermittent failures for JobTest.py on some Travis CI builds.
On my local 8 core machine, I found that playing around with the num_jobs in the JobTest.py would cause the test to fail. It was roughly (although not exactly?) related to if the num_jobs I chose was less than the number of cores available. the current value is hard coded to 11 (maybe back when 16 or 32 cores was not technically available?). Maybe this should detect the number of cores and double it? Looking at the code it seems that it could be possible that it finishes the job faster then it takes to spin up a new thread? I saw this behavior with both python 2.7 and 3.5. Can any one else test on some different number of core machines and see if they get similar results? Or does anyone have ideas what is happening here?
_______________________________________________ Scons-dev mailing list [email protected] https://pairlist2.pair.net/mailman/listinfo/scons-dev
