database/postgres
1
0
Fork 0
mirror of https://github.com/postgres/postgres.git synced 2025-11-19 13:42:17 +03:00
Code Activity

Run pgindent on 9.2 source tree in preparation for first 9.3

Browse Source 
commit-fest.
This commit is contained in:
Bruce Momjian
2012-06-10 15:20:04 -04:00
parent 60801944fa
commit 927d61eeff
494 changed files with 7343 additions and 7046 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
src/backend/storage/ipc/sinvaladt.c
View File

@@ -467,15 +467,16 @@ SIInsertDataEntries(const SharedInvalidationMessage *data, int n)
}
/*
* Now that the maxMsgNum change is globally visible, we give
* everyone a swift kick to make sure they read the newly added
* messages. Releasing SInvalWriteLock will enforce a full memory
* barrier, so these (unlocked) changes will be committed to memory
* before we exit the function.
* Now that the maxMsgNum change is globally visible, we give everyone
* a swift kick to make sure they read the newly added messages.
* Releasing SInvalWriteLock will enforce a full memory barrier, so
* these (unlocked) changes will be committed to memory before we exit
* the function.
*/
for (i = 0; i < segP->lastBackend; i++)
{
ProcState *stateP = &segP->procState[i];
stateP->hasMessages = true;
}
@@ -524,12 +525,12 @@ SIGetDataEntries(SharedInvalidationMessage *data, int datasize)
/*
* Before starting to take locks, do a quick, unlocked test to see whether
* there can possibly be anything to read. On a multiprocessor system,
* it's possible that this load could migrate backwards and occur before we
* actually enter this function, so we might miss a sinval message that
* was just added by some other processor. But they can't migrate
* backwards over a preceding lock acquisition, so it should be OK. If
* we haven't acquired a lock preventing against further relevant
* there can possibly be anything to read. On a multiprocessor system,
* it's possible that this load could migrate backwards and occur before
* we actually enter this function, so we might miss a sinval message that
* was just added by some other processor. But they can't migrate
* backwards over a preceding lock acquisition, so it should be OK. If we
* haven't acquired a lock preventing against further relevant
* invalidations, any such occurrence is not much different than if the
* invalidation had arrived slightly later in the first place.
*/