tree 50676753fde0297b6d8b92d488812ddb6893eae7
parent 3531d588c7f45c2c9c74ebeb4096595545e2730b
author Brian Norris <computersforpeace@gmail.com> 1438730964 -0700
committer Jacob Keller <jacob.e.keller@intel.com> 1438731130 -0700

aiaiai-test-patchset: sync with bisection test before moving on

For whatever reason (parallelism, I suppose?) we fork the bisectability
test to the background and continue -- we only sync with this test after
all configs are completed. This is bad for a number of reasons.

1. commit 2115f7ff0592 ("aiaiai-test-patchset: optimize disk space")
   introduced a race condition, where we might try to blow away the
   build objects before the bisection test is complete. This is really bad.

2. Theoretically, we could have more than one bisection test forked to
   the background, if there are more than 1 running config. But we
   clobber $pid_bisect every time we run test_configuration(), so we
   effectively only sync with the last one. This is *usually* OK,
   because the prior config can likely complete by the time subsequent
   config(s) complete, but that's still technically a race.

So, let's move the 'wait $pid_bisect' into the test_configuration()
function, to make sure the test is complete before we move to the next
config (and remove the build objects). While this removes a little bit
of parallelism, I suspect that parallelism was unintentional.

Signed-off-by: Brian Norris <computersforpeace@gmail.com>
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>