On Tue, Jul 30, 2019 at 04:44:11PM -0400, Tom Lane wrote:
> We do not test \h output in any existing regression test, and we're
> not going to start doing so in this one. For one thing, the expected
> URL would break every time we forked off a new release branch.
> (There would surely be value in having more-than-no test coverage
> of psql/help.c, but that's a matter for its own patch, which would
> need some thought about how to cope with instability of the output.)
One way to get out of that could be some psql-level options to control
which parts of the help output is showing up. The recent addition of
the URL may bring more weight for doing something in this area.
--
Michael