Fix the v78 migration script #5776
Conversation
Unfortunately the last fix didn't completely fix the migration to v79 of the db due to bug with schema locking during Sync2. This should fix this issue. Fix go-gitea#5759 Signed-off-by: Andrew Thornton <art27@cantab.net>
bkcsoft
added
the
lgtm/need 1
Codecov Report
@@ Coverage Diff @@
## master #5776 +/- ##
==========================================
- Coverage 37.79% 37.74% -0.05%
==========================================
Files 327 327
Lines 47840 47849 +9
==========================================
- Hits 18079 18062 -17
- Misses 27168 27194 +26
Partials 2593 2593
Continue to review full report at Codecov.
|
bkcsoft
added
lgtm/done
|
Guys before you merge if you could double check with postgres etc that this won't break them/will fix them too that would be excellent. It certainly fixes sqlite. |
|
Btw #5773 is the reason I noticed this and the reason I know that it definitely works on sqlite. |
techknowlogick
added
the
status/blocked
techknowlogick
removed
the
status/blocked
|
|
||
| // Then reset the values | ||
| sess = x.NewSession() | ||
| defer sess.Close() |
There was a problem hiding this comment.
I think this two lines is unnecessary but xorm should work.
There was a problem hiding this comment.
Hiya @lunny, dropping the index and sync2 have to be in different transactions on sqlite as the schema becomes locked by sync2 which conflicts with the index drop. Therefore the original single transaction fails on sqlite and is rolled back.
There was a problem hiding this comment.
@zeripath I know. I mean
sess := x.NewSession()
defer sess.Close()
sess.Begin()
sess.Commit()
sess.Begin()
sess.Commit()The second NewSession is uncessary.
There was a problem hiding this comment.
Oh, so xorm would just create a new transaction in the same session after the first one is committed.
Fair enough. Clean it up if you like and/or think it will cause problems in future.
|
Thanks @techknowlogick. |
Unfortunately the last fix didn't completely fix the migration to v79 of the db
due to bug with schema locking during sess.Sync2. This should fix this issue.
Fix #5759, #5779
Signed-off-by: Andrew Thornton art27@cantab.net