Horton

Why does TupleDescInitBuiltinEntry lack a "default: error" case?

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: pgsql-hackers@lists.postgresql.org
Date: 2018-12-09T19:49:35Z
Lists: pgsql-hackers 
If the passed-in type OID isn't one of the very short list that
TupleDescInitBuiltinEntry supports, it will silently hand back
a broken TupleDesc, rather than throwing an error.  How can
this possibly be considered good code?

(So far as I can see, none of the extant callers could hit
such an error, but it's still bad code.)

			regards, tom lane

Commits

  1. Make TupleDescInitBuiltinEntry throw error for unsupported types.