OSDN Git Service

Remove an unnecessary HOLD_INTERRUPTS/RESUME_INTERRUPTS pair.
authorTom Lane <tgl@sss.pgh.pa.us>
Sun, 22 Oct 2006 20:34:54 +0000 (20:34 +0000)
committerTom Lane <tgl@sss.pgh.pa.us>
Sun, 22 Oct 2006 20:34:54 +0000 (20:34 +0000)
This was required back when RESUME_INTERRUPTS could actually
execute ProcessInterrupts, but that hasn't been true since 2001...

src/backend/storage/buffer/bufmgr.c

index 5efd9cc..45cdace 100644 (file)
@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *       $PostgreSQL: pgsql/src/backend/storage/buffer/bufmgr.c,v 1.212 2006/10/04 00:29:57 momjian Exp $
+ *       $PostgreSQL: pgsql/src/backend/storage/buffer/bufmgr.c,v 1.213 2006/10/22 20:34:54 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -694,21 +694,14 @@ retry:
                BufTableDelete(&oldTag, oldHash);
 
        /*
-        * Avoid accepting a cancel interrupt when we release the mapping lock;
-        * that would leave the buffer free but not on the freelist.  (Which would
-        * not be fatal, since it'd get picked up again by the clock scanning
-        * code, but we'd rather be sure it gets to the freelist.)
+        * Done with mapping lock.
         */
-       HOLD_INTERRUPTS();
-
        LWLockRelease(oldPartitionLock);
 
        /*
         * Insert the buffer at the head of the list of free buffers.
         */
        StrategyFreeBuffer(buf, true);
-
-       RESUME_INTERRUPTS();
 }
 
 /*