PIFT creating duplicate test records, causing PDO exception during pift-cron [#2126225]

After the D7 upgrade, pift-cron is crashing due to PDO duplicate key issues during a db_update of pift_data.

This is likely coming from pift_test_sent(), which updates test request records with a test_id, after qa.d.o has assigned one for that test. The code assumes we have a unique identifier for test_id; but we are somehow getting a number of tests with '0' instead.

I suspect this may be occuring when a file is uploaded, the node is saved, and then it is saved and edited again before pift-cron has assigned the file a test_id. Either that, or the pift_test_add_files logic is broken ... but somehow tests are being double-queued.

Comments

FWIW, it happened to me on #2019055: Switch from field-level language fallback to entity-level language fallback, in comment #168 where I changed the issue to RTBC. Upon submitting the form, the next page displayed a sh*tload of "patch 2019055-xx.patch set to be retested" confirmation messages (one per patch in the issue, which was a *lot*).

Comment #4

jthorson CreditAttribution: jthorson commented 4 November 2013 at 03:50

yched: And thanks for doing that ... because it wasn't until I realized all 46 duplicate records were from that same issue that I was able to deduce what was actually going on. :D