Hannes Dorbath <light@theendofthetunnel.de> writes:
> Alvaro Herrera wrote:
>> Can you please provide a backtrace from gdb?
> I hope that contains it: http://theendofthetunnel.de/backtrace.log
Hmmm --- one thing that jumps out at me is that SplitToVariants assumes
(in four places) that the SplitVar.stem arrays can't overflow. Now 1024
entries does sound like plenty, but perhaps a malformed dictionary could
cause the code to overrun that? Or an extremely long input word?
Personally I'd put some checks on nstem in there ...
regards, tom lane