Mayeul Kauffmann <mayeul.kauffmann@free.fr> writes:
> I am trying to patch the server source to increase the number of columns
> above 1600. I'm not planning to commit this but as suggested elsewhere
> [1], someone might suggest a configure option based on this.
> I came up with a patch which seems to work (see below), but 3 of the 136
> tests fail.
You would have to show us the actual failure diffs to get much useful
comment, but in general increasing the size of tuple headers could
easily lead to changes in plan choices, which would affect output
row ordering (not to mention EXPLAIN results). This is particularly
the case if you're testing on a 64-bit machine, since the maxalign'd
size of the header would go from 24 to 32 bytes ...
regards, tom lane