Robert Haas <robertmhaas@gmail.com> writes:
> On Thu, Aug 17, 2017 at 2:06 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Nope, spoke too soon. See buildfarm.
> Whoa, that's not good.
Ah-hah, I see my dromedary box is one of the ones failing, so I'll
have a look there. I can't reproduce it on my other machines.
I'm a bit suspicious that it's got something to do with getting
a different number of workers during restart. Whether that's
the issue or not, though, it sure seems like a rescan leaks an
unpleasantly large amount of memory. I wonder if we shouldn't
refactor this so that the per-reader structs can be reused.
regards, tom lane
