christoph.berg@credativ.de writes:
> This seems to be a bug:
> time psql -c "SET statement_timeout = '3s'; SELECT pg_sleep(2)"
> ERROR: canceling statement due to statement timeout
> real 0m1.065s
The reason this isn't a bug is that a -c command string is sent to the
server as a single statement (PQexec call), and what "statement timeout"
controls is the total time for the whole thing. The SET operation can't
change the already-running timer for the current statement. It would
affect the timeout for the next statement ... but there won't be one.
Many people have complained that it's unintuitive that -c works this way
rather than breaking up the string into multiple submissions the same way
psql would do with normal input. We're afraid to change it for fear of
breaking applications, though. If you want behavior more like psql's
normal operation, consider
echo "SET statement_timeout = '3s'; SELECT pg_sleep(2)" | psql
regards, tom lane
