- Notifications
You must be signed in to change notification settings - Fork5
Commitde396a1
committed
Properly re-initialize replication slot shared memory upon creation.
Slot creation did not clear all fields upon creation. After start thememory is zeroed, but when a physical replication slot was created inthe shared memory of a previously existing logical slot, catalog_xminwould not be cleared. That in turn would prevent vacuum from doing itsduties.To fix initialize all the fields. To make similar future bugs lesslikely, zero all of ReplicationSlotPersistentData, and re-order therest of the initialization to be in struct member order.Analysis: Andrew GierthReported-By: md@chewy.comAuthor: Michael PaquierDiscussion: <20160705173502.1398.70934@wrigleys.postgresql.org>Backpatch: 9.4, where replication slots were introduced1 parent509815e commitde396a1
1 file changed
+14
-4
lines changedLines changed: 14 additions & 4 deletions
Original file line number | Diff line number | Diff line change | |
---|---|---|---|
| |||
263 | 263 |
| |
264 | 264 |
| |
265 | 265 |
| |
266 |
| - | |
267 |
| - | |
268 |
| - | |
| 266 | + | |
| 267 | + | |
| 268 | + | |
269 | 269 |
| |
270 | 270 |
| |
271 |
| - | |
| 271 | + | |
| 272 | + | |
| 273 | + | |
| 274 | + | |
| 275 | + | |
| 276 | + | |
| 277 | + | |
| 278 | + | |
| 279 | + | |
| 280 | + | |
| 281 | + | |
272 | 282 |
| |
273 | 283 |
| |
274 | 284 |
| |
|
0 commit comments
Comments
(0)