Front page | perl.perl5.porters |
Postings from February 2013
From: Nicholas Clark
February 28, 2013 19:16
Message ID: 20130228191638.GA3729@plum.flirble.org
On Thu, Feb 28, 2013 at 06:14:24PM +0100, Steffen Mueller wrote:
> On 02/28/2013 06:01 PM, Andy Dougherty wrote:
> > What about CPAN modules using it? There are at least a few. I forget
> > how we deal with such things.
> Ah, good catch. I don't know how much interpreter variables are
> considered to be public API.
The "API" was only ever really defined* for C functions.
Variables, data structures and C macros seem to be treated as fair game.
> If in doubt, the non-cost of simply keeping the variable, ignoring it in
> the actual core code, adding a deprecation to the perldelta, and then
> dropping it in 5.20 seems like the right trade-off here.
Yes, this seems best.
Any code that doesn't get the message in advance will fail to compile with
the first dev release that removes it. Which is good, as compile-time failure
is the least worst failure mode.
* pretty much as "list the existing functions that seem to be useful to XS