514504 - Gloda: messages for which the clean bit hasn't been committed yet can be indexed in both event and sweep modes

Status: RESOLVED FIXED

(MailNews Core :: Database, defect)

Description

[Rain [:rain1] (inactive)]

STR:

1. Trigger a large number of msgAdded notifications -- around 40 or so should do.
2. We'll start indexing the first few messages in event mode, but soon go into sweep mode.
3. Once in sweep mode, we'll attempt to index the same messages again.

This happens because we only set the clean bit on the message header after we've committed to disk, and we never get a chance in between steps 2 and 3 to commit to disk. Once we're in sweep mode, we completely forget that we've already indexed these messages.

test_index_*_imap_messages_in_bulk.js should be able to test this.

Comment 1

[Wayne Mery (:wsmwk)]

Attachment: activity mgr & folder with glodazilla info

to go with the screenshot - activity mgr says 11 indexed, there are 7 msgs in the folder, 2 of which were moved <1/2 hr before

unclear if what I saw was this bug.

Comment 2

[Andrew Sutherland [:asuth] (he/him)]

The gloda correctness patch addresses this through the introduction of the PendingCommitTracker.  This is tested in base_index_messages.js by the aptly named test_sweep_indexing_does_not_reindex_event_indexed.

Comment 3

[Andrew Sutherland [:asuth] (he/him)]

Fixed in gloda correctness patch tracked on bug 465618.
trunk commit: http://hg.mozilla.org/comm-central/rev/413b2018349c