I wrote:
> David Fetter <david@fetter.org> writes:
>> [ v7-0001-Add-warn-to-psql.patch ]
> I took a look at this. I have no quibble with the proposed feature,
> and the implementation is certainly simple enough. But I'm unconvinced
> about the proposed test scaffolding.
I pushed this with the simplified test methodology.
While I was fooling with it I noticed that the existing code for -n
is buggy. The documentation says clearly that only the first
argument is a candidate to be -n:
If the first argument is an unquoted <literal>-n</literal> the trailing
newline is not written.
but the actual implementation allows any argument to be recognized as
-n:
regression=# \echo this -n should not be -n like this
this should not be like thisregression=#
I fixed that, but I'm wondering if we should back-patch that fix
or leave the back branches alone.
regards, tom lane