NunoACHenriques <nach(at)fct(dot)unl(dot)pt> writes:
> A different error today:
> [MemoryContextAlloc: invalid request size 4294967295]
This could be a variant of the same problem: instead of getting a zero
tuple length from the sort temp file, we're reading a -1 tuple length.
Still no way to tell if it's a hardware glitch or a software bug.
(If the latter, presumably the code is getting out of step about its
read position in the temp file --- but how?)
regards, tom lane