fb5d54d506
Instead of using the zake provided flush method (which does have issues, since it does not guarantee that the associated watches will have been called) instead use a new method which creates nodes, waits for there watches to be triggered, does other work, and then deletes the node and waits for that watcher to be triggered; this is more stable (as it depends on the linearity guarantee of zookeeper and the kazoo threading model). Change-Id: I12fd9c7bcc5cd9009b4175166edfc924e94161bf
12 lines
178 B
Plaintext
12 lines
178 B
Plaintext
hacking>=0.8.0,<0.9
|
|
discover
|
|
coverage>=3.6
|
|
mock>=1.0
|
|
python-subunit>=0.0.18
|
|
testrepository>=0.0.18
|
|
testtools>=0.9.34
|
|
zake>=0.0.18
|
|
# docs build jobs
|
|
sphinx>=1.1.2,<1.2
|
|
oslosphinx
|