Thread: pgstattuple does not contain CHECK_FOR_INTERRUPTS()

Hi,

pgstattuple (and pgstatindex) does not contain CHECK_FOR_INTERRUPTS().
Therefore, we can not stop pgstattuple() by using the signal while a
large table is processed.

Here is the patch to solve it.

Best regards,

--
NTT OSS Center
Tatsuhito Kasahara
diff -cr pgsql/contrib/pgstattuple/pgstattuple.c pgsql_work/contrib/pgstattuple/pgstattuple.c
*** pgsql/contrib/pgstattuple/pgstattuple.c    2009-06-11 23:48:52.000000000 +0900
--- pgsql_work/contrib/pgstattuple/pgstattuple.c    2010-04-02 18:52:05.000000000 +0900
***************
*** 276,281 ****
--- 276,283 ----
      /* scan the relation */
      while ((tuple = heap_getnext(scan, ForwardScanDirection)) != NULL)
      {
+         CHECK_FOR_INTERRUPTS();
+
          /* must hold a buffer lock to call HeapTupleSatisfiesVisibility */
          LockBuffer(scan->rs_cbuf, BUFFER_LOCK_SHARE);

***************
*** 302,307 ****
--- 304,311 ----

          while (block <= tupblock)
          {
+             CHECK_FOR_INTERRUPTS();
+
              buffer = ReadBuffer(rel, block);
              LockBuffer(buffer, BUFFER_LOCK_SHARE);
              stat.free_space += PageGetHeapFreeSpace((Page) BufferGetPage(buffer));
***************
*** 313,318 ****
--- 317,324 ----

      while (block < nblocks)
      {
+         CHECK_FOR_INTERRUPTS();
+
          buffer = ReadBuffer(rel, block);
          LockBuffer(buffer, BUFFER_LOCK_SHARE);
          stat.free_space += PageGetHeapFreeSpace((Page) BufferGetPage(buffer));
***************
*** 454,459 ****
--- 460,467 ----
      blkno = start;
      for (;;)
      {
+         CHECK_FOR_INTERRUPTS();
+
          /* Get the current relation length */
          LockRelationForExtension(rel, ExclusiveLock);
          nblocks = RelationGetNumberOfBlocks(rel);

Tatsuhito Kasahara <kasahara.tatsuhito@oss.ntt.co.jp> writes:
> pgstattuple (and pgstatindex) does not contain CHECK_FOR_INTERRUPTS().
> Therefore, we can not stop pgstattuple() by using the signal while a
> large table is processed.

> Here is the patch to solve it.

Seems to be a good idea --- will apply.  Thanks!
        regards, tom lane