Re: Remove hardcoded hash opclass function signature exceptions

Peter Eisentraut <peter(at)eisentraut(dot)org> writes:
> hashvalidate(), which validates the signatures of support functions for
> the hash AM, contains several hardcoded exceptions.
> ...
> This patch removes those exceptions by providing new support functions
> that have the proper declared signatures. They internally share most of
> the C code with the "wrong" functions they replace, so the behavior is
> still the same.

+1 for cleaning this up. A couple of minor nitpicks:

* I don't really like the new control structure, or rather lack of
structure, in hashvalidate. In particular the uncommented
s/break/continue/ changes look like typos. They aren't, but can't
you do this in a less confusing fashion? Or at least add comments
like "continue not break because the code below the switch doesn't
apply to this case".

* Hand-picking OIDs as you did in pg_proc.dat is kind of passé now.
I guess it's all right as long as nobody else does the same thing in
the near future, but ...

> Not done here, but maybe hashvarlena() and hashvarlenaextended() should
> be removed from pg_proc.dat, since their use as opclass support
> functions is now dubious.

I wish we could get rid of those, but according to
codesearch.debian.net, postgis and a couple of other extensions are
relying on them. If we remove them we'll break any convenient upgrade
path for those extensions.

regards, tom lane