> On Sat, Oct 20, 2012 at 9:49 PM, Daniel Farina <daniel@heroku.com> wrote:
>>
>> An unrelated defect, although the patch tries to carefully clean up
>> the 'res' result from psqlexec in the error cases, it does forget to
>> do that, seemingly, in the 'positive' case, while it is looping. I
>> think it needs another pqclear in there.
>
I have another unrelated defect. I think the lower bound of "zero"
time for repeating a query is too low -- I made a bad query and ran it
against my local machine and my terminal was flooded with some kind of
output (maybe a NULL byte per iteration?). I couldn't look too
carefully, because it basically hosed that terminal session almost
instantly. Also, because the interrupts are enabled only during the
sleep section (this may also be a defect), I couldn't send SIGINT very
well. I think. My computer was struggling, in any case.
Deceptively, this would not occur on any database that was remote,
because you'd get at *least* a millisecond of delay.
--
fdr