]> granicus.if.org Git - postgresql/commit
Fix a many-legged critter reported by chifungfan@yahoo.com: under the
authorTom Lane <tgl@sss.pgh.pa.us>
Tue, 22 Aug 2000 04:06:22 +0000 (04:06 +0000)
committerTom Lane <tgl@sss.pgh.pa.us>
Tue, 22 Aug 2000 04:06:22 +0000 (04:06 +0000)
commit0147b1934f251183d3614bca011bf21205890835
treeed7df11ba0ecbdae22095a2eeacbd204dcdca1b8
parent94e90d9a86a186c83891fe4ce3e343bcf1860053
Fix a many-legged critter reported by chifungfan@yahoo.com: under the
right circumstances a hash join executed as a DECLARE CURSOR/FETCH
query would crash the backend.  Problem as seen in current sources was
that the hash tables were stored in a context that was a child of
TransactionCommandContext, which got zapped at completion of the FETCH
command --- but cursor cleanup executed at COMMIT expected the tables
to still be valid.  I haven't chased down the details as seen in 7.0.*
but I'm sure it's the same general problem.
src/backend/commands/copy.c
src/backend/executor/execMain.c
src/backend/executor/execUtils.c
src/backend/executor/nodeHash.c
src/backend/tcop/pquery.c
src/include/executor/hashjoin.h
src/include/nodes/execnodes.h