"Rod Taylor" <rbt@zort.ca> writes:
> Anyway, I copped out and added a %expect 1.
Sorry, but we agreed long ago that %expect is *not* an acceptable
alternative to getting the bugs out of your yacc productions.
%expect is okay for stable grammars or ones that are being hacked by
just one or two closely-involved people. The PG grammar is being hit on
all the time by a lot of people with varying yacc skill, and some of the
rearrangements are not trivial. The problem with %expect for us is: all
right, we just ignored N shift/reduce errors --- but are they the same N
errors that someone once long ago decided were okay? It's too risky.
I believe your real problem has nothing to do with the c_expr anyway,
but is that DEFAULT is an allowable ColLabel. I do not recall which
keywords we currently need to accept as def_elem labels, but perhaps
it'd work to back off the lefthand side from ColLabel to ColId or
some such.
regards, tom lane