On Wed, May 17, 2017 at 12:47 AM, Robert Haas <robertmhaas@gmail.com> wrote:
> On Mon, May 15, 2017 at 9:59 PM, Michael Paquier
> <michael.paquier@gmail.com> wrote:
>> + *
>> + * Returns -1 on failure, 0 if the socket is readable/writable, 1 if
>> it timed out.
>> */
>> pqWait is internal to libpq, so we are free to set up what we want
>> here. Still I think that we should be consistent with what
>> pqSocketCheck returns:
>> * >0 means that the socket is readable/writable, counting things.
>> * 0 is for timeout.
>> * -1 on failure.
>
> That would imply a lot more change, though. The way that the patch
> currently does it, none of the other callers of pqWait() or
> pqWaitTimed() need to be adjusted. So I prefer the way that Tsunakawa
> Takayuki currently has this over your proposal.
Consistency in APIs matters, but I won't fight hard in favor of this
item either. In short I am fine to discard this comment.
--
Michael
