Tomas Vondra <[email protected]> writes: > I wonder if we could check the index tuple length, and adjust the siglen > based on that, somehow ...
In an already-corrupted index, there won't be a unique answer.
regards, tom lane
