Tom Lane wrote:
>
> Hiroshi Inoue <Inoue@tpf.co.jp> writes:
> > Does this change also affect the SIGTERM handling
> > (not only query-cancel) ?
>
> Yes. Cancel and die interrupts are accepted at the same places.
Cancel interrupts are accepted at CHECK_FOR_INTERRUPTS
or while waiting for a lock.
Die interrupts are accepted at CHECK_FOR_INTERRUPTS or
at the time when they are received while interrupts
are allowed. Is my understanding right ? If so, isn't
it strange ?
regards,
Hiroshi Inoue