Make sure that open hash table scans are cleaned up when bgwriter tries to
authorTom Lane <tgl@sss.pgh.pa.us>
Tue, 11 Sep 2007 17:15:55 +0000 (17:15 +0000)
committerTom Lane <tgl@sss.pgh.pa.us>
Tue, 11 Sep 2007 17:15:55 +0000 (17:15 +0000)
commit5e6378e220cbfd02aed5b8e7aacc8a82c6e91d4b
tree08cb10b2b5a199ce82d30b0db0811d8874180254
parent125603f4bd009dba5348be4b976cd7a13c7da294
Make sure that open hash table scans are cleaned up when bgwriter tries to
recover from elog(ERROR).  Problem was created by introduction of hash seq
search tracking awhile back, and affects all branches that have bgwriter;
in HEAD the disease has snuck into autovacuum and walwriter too.  (Not sure
that the latter two use hash_seq_search at the moment, but surely they might
someday.)  Per report from Sergey Koposov.
src/backend/postmaster/autovacuum.c
src/backend/postmaster/bgwriter.c
src/backend/postmaster/walwriter.c