develooper Front page | perl.perl5.porters | Postings from October 2015

Re: patches wanted: default warnings/strict

Thread Previous | Thread Next
Aristotle Pagaltzis
October 19, 2015 12:14
Re: patches wanted: default warnings/strict
Message ID:
* Ricardo Signes <> [2015-10-19 04:40]:
> If we admit that we may want new warnings to be available, but not
> turned on by "use warnings;", and that they can only be added to the
> core, then it is going to be the case that the user will need to add
> something to request them. This strategy just avoids the needless
> multiplication of interfaces to do that.

Sure, such as using Perl::Critic on their code. And obviously there is
runtime stuff that Perl::Critic cannot catch, that it would be nice to
be able to catch via some mechanism. That is the requirement, and it is
a reasonable one.

Implementing those checks as warnings, which `use warnings` does not
enable, is *a* solution for this requirement, but not the only possible
one, and I remain unconvinced it is even a good one, let alone the best
(or something in the ballpark).

I really don’t feel comfortable with the “well it seems reasonable, and
Ævar really wants it (by reasoning from a bad analogy with C compilers),
so let’s proceed” approach. That’s how we got autoderef, except this one
would be far harder to excise if it turns out to be a mistake; more like
v-strings, a regret to live with forever after.

Can we somehow make this an experimental feature? I am willing to have
my mind changed by an absence of problems in practice, but would prefer
if everyone else also received an opportunity to have their mind changed
by the presence of problems in practice.

(Of course there is also the problem that once this makes it out to the
real world, it will incur breakage – and a potentially terrible amount
thereof. That would be a huge waste of goodwill if the feature turns out
to be regrettable. So the initial incarnation in blead would probably
have to be non-experimental, and CPAN-smoked, to get an idea of how much
terror to expect.)

Aristotle Pagaltzis // <>

Thread Previous | Thread Next Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at | Group listing | About