On Fri, Oct 7, 2011 at 9:59 PM, Bruce Momjian <bruce@momjian.us> wrote:
> Robert Haas wrote:
>> On Fri, Oct 7, 2011 at 4:06 PM, Bruce Momjian <bruce@momjian.us> wrote:
>> > Robert Haas wrote:
>> >> On Fri, Oct 7, 2011 at 1:03 PM, Kevin Grittner
>> >> <Kevin.Grittner@wicourts.gov> wrote:
>> >> > Robert Haas <robertmhaas@gmail.com> wrote:
>> >> >> The funny thing is that I've been thinking all of these months
>> >> >> about how convenient it is that we defined WAL_DEBUG in debug
>> >> >> builds
>> >> >
>> >> > IMO, --enable-debug should not do anything but include debugging
>> >> > symbols. ?The ability to get a useful stack trace from a production
>> >> > crash, without compromising performance, is just too important by
>> >> > itself to consider conditioning any other behavior on it.
>> >>
>> >> So, should I go revert this change in head and 9.1, or does anyone
>> >> else want to argue for Heikki's position that we should just leave it
>> >> on, on the theory that it's too cheap to matter?
>> >
>> > I would just fix it in head.
>>
>> That just seems weird. Either it's cheap enough not to matter (in
>> which case there's no reason to revert that change at all) or it's
>> expensive enough to matter (in which case presumably we don't want to
>> leave it on in 9.1 for the 5 years or so it remains a supported
>> release).
>
> I am concerned about changing behavior on people in a minor release ---
> it is not about risk in this case.
Well, I still maintain that if the performance impact is low enough
that we can get away with that, it's probably not worth fixing in
master either. But at any rate, we now have three opinions on what to
do about this. Anyone else want to cast a vote (preferably not for an
entirely new option)?
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company