Re: plpgsql fails to reinitialize record variables at block re-entry

Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com> writes:
> 2017-12-09 7:24 GMT+01:00 Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>:
>> Surely that ought to have failed at the i=2 iteration. There is
>> code in plpgsql's exec_stmt_block() that tries to reinitialize
>> PLPGSQL_DTYPE_REC variables, but it's never reached (as a quick
>> look at coverage.postgresql.org will confirm), because what it
>> scans is only the variables attached to the block by
>> plpgsql_add_initdatums() --- and that function thinks it should
>> only pay attention to PLPGSQL_DTYPE_VAR variables.

> +1

Pushed. Some excavation in our git history says this bug has been
there since the initial commit of plpgsql, making it a little over
19 years old. Might be a new record for the age of a PG bug.

regards, tom lane