In master I've been testing some new code that I'm working on around foreign keys. I wasn't quite sure if it was possible to include the same column twice in a foreign key, so I tested....
create table r1 (a int); create table r2 (b int); create unique index on r2(b,b); alter table r1 add constraint r2_b_fkey foreign key (a,a) references r2 (b,b); ERROR: cache lookup failed for opclass 0 Which is not quite any of the messages that I would have expected I've tracked this down to a small bug in transformFkeyCheckAttrs() where opclasses[1] won't get set because if (attnums[0] == indexStruct->indkey.values[1]) will match then break out of the inner loop. Instead opclasses[0] gets set twice. The attached seems to fix the problem, but the whole thing makes me wonder if this is even meant to be allowed? I was thinking that this might be a good time to disallow this altogether, since it's already broken and looks like it has been for about 11 years Disallowing it would simplify some code in my semi/anti join removal patch that I posted here http://www.postgresql.org/message-id/CAApHDvpCBEfuc5tD=vniepAv0pU5m=q=foqzcodmheei7oq...@mail.gmail.com Any thoughts? Regards David Rowley
transformFkeyCheckAttrs_fix.patch
Description: Binary data
-- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
