Example: T132804 was removed from Q4 project in April but still shows up in Q4 in Phlogiston report as of August.
Another example: T96061.
Description
Description
Related Objects
Related Objects
Event Timeline
Comment Actions
Notes from debugging:
The raw data for which projects each task should belong to on each day is present in Phlogiston's loaded dump:
phab=# select date_modified, active_projects, has_edge_data
from maniphest_transaction
where task_id = 132804
and transaction_type = 'core:edge';
date_modified | active_projects | has_edge_data
------------------------+-----------------+---------------
2016-04-15 17:43:12+00 | {1922,1903} | t
2016-04-15 17:43:12+00 | {} | f
2016-04-15 18:24:03+00 | {} | f
2016-04-15 18:26:50+00 | {} | f
2016-04-29 15:40:11+00 | {1922,65} | t
2016-08-05 04:20:33+00 | {} | f
(6 rows)But the edge data reconstructed from this is clearly incorrect:
phab=# select * from maniphest_edge where task = 132804 and edge_date > '2016-04-27'; task | project | edge_date --------+---------+------------ 132804 | 1903 | 2016-04-28 132804 | 1922 | 2016-04-28 132804 | 1903 | 2016-04-29 132804 | 1922 | 2016-04-29 132804 | 1922 | 2016-04-30 132804 | 1903 | 2016-04-30 132804 | 1922 | 2016-05-01 132804 | 1903 | 2016-05-01 132804 | 1922 | 2016-05-02 132804 | 1903 | 2016-05-02
Comment Actions
Notes from debugging: fixed that problem by restoring the "LIMIT 1" line in the stored procedure. The task_history table is now correct. task_history_recat, however, is incorrect, and the reason is that apply_tag_based_recategorization isn't constrained by date.
Comment Actions
I think this is now fixed. The two test tasks are both "Other". Please confirm, maybe check some other tasks.