If large results values are returned by the subprocesses, we can hit a deadlock where the subprocess is trying to write data back to the parent, the pipe is full and the parent is waiting for the child to exit.
Avoid this by calling the update() method which would trigger reading a result from the child, avoiding the deadlock. The issue is described in https://bugs.python.org/issue8426 Signed-off-by: Richard Purdie <[email protected]> --- meta/lib/oe/utils.py | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/meta/lib/oe/utils.py b/meta/lib/oe/utils.py index 0c1d48a209e..3a496090f3a 100644 --- a/meta/lib/oe/utils.py +++ b/meta/lib/oe/utils.py @@ -307,6 +307,10 @@ def multiprocess_launch(target, items, d, extraargs=None): p.start() launched.append(p) for q in launched: + # Have to manually call update() to avoid deadlocks. The pipe can be full and + # transfer stalled until we try and read the results object but the subprocess won't exit + # as it still has data to write (https://bugs.python.org/issue8426) + q.update() # The finished processes are joined when calling is_alive() if not q.is_alive(): if q.exception: -- 2.20.1 -- _______________________________________________ Openembedded-core mailing list [email protected] http://lists.openembedded.org/mailman/listinfo/openembedded-core
