... btw, Coverity also doesn't like this fragment of the patch: /srv/coverity/git/pgsql-git/postgresql/src/bin/psql/common.c: 1084 in ShowNoticeMessage() 1078 static void 1079 ShowNoticeMessage(t_notice_messages *notes) 1080 { 1081 PQExpBufferData *current = notes->in_flip ? ¬es->flip : ¬es->flop; 1082 if (current->data != NULL && *current->data != '\0') 1083 pg_log_info("%s", current->data); >>> CID 1476041: Null pointer dereferences (FORWARD_NULL) >>> Passing "current" to "resetPQExpBuffer", which dereferences null >>> "current->data". 1084 resetPQExpBuffer(current); 1085 } 1086 1087 /* 1088 * SendQueryAndProcessResults: utility function for use by SendQuery() 1089 * and PSQLexecWatch().
Its point here is that either the test of "current->data != NULL" is useless, or resetPQExpBuffer needs such a test too. I'm inclined to guess the former. (Just as a matter of style, I don't care for the flip/flop terminology here, not least because it's not clear why exactly two buffers suffice and will suffice forevermore. I'd be inclined to use an array of two buffers with an index variable.) regards, tom lane