]> granicus.if.org Git - postgresql/commit
Found the problem with my operator-family changes: by fetching from
authorTom Lane <tgl@sss.pgh.pa.us>
Sun, 31 Dec 2006 20:32:04 +0000 (20:32 +0000)
committerTom Lane <tgl@sss.pgh.pa.us>
Sun, 31 Dec 2006 20:32:04 +0000 (20:32 +0000)
commit0b56be83441c01419fcf82ebe666e968e6f7b246
tree77aafcc3cb26fcdc5284844227cedf88c6f7957a
parent5725b9d9afc8c3ba24e94cbc7020889fe8ad7ef9
Found the problem with my operator-family changes: by fetching from
pg_opclass during LookupOpclassInfo(), I'd turned pg_opclass_oid_index
into a critical system index.  However the problem could only manifest
during a backend's first attempt to load opclass data, and then only
if it had successfully loaded pg_internal.init and subsequently received
a relcache flush; which made it impossible to reproduce in sequential
tests and darn hard even in parallel tests.  Memo to self: when
exercising cache flush scenarios, must disable LookupOpclassInfo's
internal cache too.
src/backend/utils/cache/relcache.c