Every time the setup function runs, it chooses a random page name with
random content. If it chooses a page name + content combination that was
used to set up a different test in the same run, the edit it makes will
be a null edit, the revid will be null, and the test will crash.
There are 100 possible page names and 100 possible page texts, so 10k
possibilities total, and the setup function appears to be called 10 times,
so the probability of a collision would have been
(10 choose 2)/10000 = 0.0045 = 0.45%, or one in every 222 times.
Bug: T151878
Change-Id: I800ee8512c2ad171ba793bea343f456653f9a16d
Originally introduced in b84eedc74e, it was reverted for security concerns.
New changes:
* Instead of bundling the log summary with notification, load it on display
* If the log event has been suppressed after the thanks for it has been sent,
silently delete the event to prevent the confusion of linking to something
zapped.
* Keep the 'already sent' cache key compatible with old format
* Validate the log id in the API
* Change ApiCoreThank::getRevisionFromParams() to ApiCoreThank::getRevisionFromId()
Bug: T186855
Bug: T188791
Depends-On: Ic5e9db0def857d9dcecbd06bf081c8c83712c1ea
Change-Id: I03aea7d9f4dfa0fe49639c53968deabf89999d2d
Add a 'log' API parameter (for the log ID to thank), and add
a whitelist config variable for specifying which log types are
thankable.
Bug: T186855
Change-Id: I58ae90c9729c0066f952e90fca2cf99b029d0d9b
