Tom Lane wrote:
>
> Hiroshi Inoue <Inoue@tpf.co.jp> writes:
> > When are cancel or die interrupts accepted while
> > executing a long query ?
>
> CHECK_FOR_INTERRUPTS() calls. The call in ExecProcNode should alone
> suffice for most queries, I think.
I couldn't cancel VACUUM using current cvs.
I couldn't cancel simple SELECT which returns no row either.
...
IIRC cancel interrupts before 7.1 was so but die interrupts
has never been so.
regards,
Hiroshi Inoue