]> granicus.if.org Git - postgresql/commit
Fix canAcceptConnections() bugs introduced by replication-related patches.
authorTom Lane <tgl@sss.pgh.pa.us>
Sun, 14 Nov 2010 20:57:42 +0000 (15:57 -0500)
committerTom Lane <tgl@sss.pgh.pa.us>
Sun, 14 Nov 2010 20:57:42 +0000 (15:57 -0500)
commitc350b01e2f72249ce6bcb642e0ee6730838b0cfe
tree046e56598fff74e4879188952d127fcaaeb4ffab
parent9f65a874ebf86291ca1f7920f74f6c9d6d20a213
Fix canAcceptConnections() bugs introduced by replication-related patches.

We must not return any "okay to proceed" result code without having checked
for too many children, else we might fail later on when trying to add the
new child to one of the per-child state arrays.  It's not clear whether
this oversight explains Stefan Kaltenbrunner's recent report, but it could
certainly produce a similar symptom.

Back-patch to 8.4; the logic was not broken before that.
src/backend/postmaster/postmaster.c