I wrote:
> Manuel Kniep <manuel@adjust.com> writes:
>> ok after lot’s of testing I could create a test case
>> which can be found here https://gist.github.com/rapimo/3c8c1b35270e5854c524
>> it’s written in ruby an depends on the gem activerecord pg and parallel
> Hm. I don't see a segfault from this. I do see the CREATE TEMP TABLE
> command failing with "ctid is NULL", which probably shouldn't be happening
> ... but no segfault.
The reason turns out to be that this is a dangling-pointer bug, and I was
using a memory-clobber-enabled build so it was pretty predictable what the
pointer would be pointing at. I've got no doubt that hard-to-reproduce
misbehavior, including segfaults, would ensue without CLOBBER_FREED_MEMORY
turned on.
You need this patch:
http://git.postgresql.org/gitweb/?p=postgresql.git;a=patch;h=34668c8eca065d745bf1166a92c9efc588e7aee2
regards, tom lane
