On 30 December 2012 04:37, Robert Haas <robertmhaas@gmail.com> wrote:
> On Sat, Dec 29, 2012 at 10:07 PM, Greg Smith <greg@2ndquadrant.com> wrote:
>> It is a strange power of two to be appearing there. I can follow your
>> reasoning for why this could be a bit flipping error. There's no sign of
>> that elsewhere though, no other crashes under load. I'm using this server
>> here because it's worked fine for a while now.
>
> In theory, if this is sufficiently reproducible, could you start
> backing up through the commit history 100 or so commits at a time
> until you find the culprit? Or use git bisect? It'd be kind of
> time-consuming, but...
FYI, Andrew Dunstan recently blogged about using git bisect to find
problems in Postgres. His example shell script will save you the
trouble of independently coming up with much the same boilerplate (the
non-boilerplate parts are probably what you need to worry about,
though).
http://people.planetpostgresql.org/andrew/index.php?/archives/301-Finding-the-origin-of-problems-with-git-bisect.html
If you try and do a binary search through many commits *manually*, for
a bug that is this hard to reproduce, well, that's no fun.
--
Peter Geoghegan http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training and Services