Bug 357655

Description Chris Recoskie 2011-09-14 12:08:18 EDT

Similar problem to Bug 357653

I've run into this problem... this is an over-simplification of the scenario
(the local builders do not exhibit this problem), but it illustrates the issue.


- Remote make builder creates an RSEProcess to do the build and launches a thread to keep tabs on it.  If this thread notices that the progress monitor is canceled, it will try to destroy the process

- the builder then does a waitFor() on the make process to know when the build
is done, and update the progress accordingly

- user cancels the progress monitor, but RSEProcess.destroy() blocks because the builder holds the lock on the process via its call to waitFor()

I don't think destroy() should ever by synchronized.  You should be able to
kill a given process no matter what state it's in;  it's your last resort
against something that's running amok.  Having it synchronized means you are
forced to wait for it to complete if anyone anywhere is waiting for it to
complete.