perl5.20

[Daniel J. Luke]

On Jul 22, 2014, at 5:25 PM, David Evans <devans at macports.org> wrote:
> 
> On 7/22/14 1:01 PM, Daniel J. Luke wrote:
>>> I think it is way too early to be switching to it by default. Much less the only option. So I vote we at least keep perl 5.16 and perl 5.18 and all of their modules around for a while.  
>> only perl5.18 currently has a maintainer in MacPorts.
>> 
>> upstream has perl5.18 in maintenance mode, and recommends installing perl5.20 (previous perl5.x versions are all End of life).
> We have two topics going here:
> 
>  1) trimming old/outdated/unsupported perl ports
>  2) what to do about Mojca's commit that broke perl5.20 builds
> 
> I suggest we do the following:
> 
>  * agree on which old perl5 ports can be dispensed with.  Looks like we
> can agree at this point on removing anything older than perl5.16 as a
> starting point.  I tend to support inclusion of perl5.16 for now because
> it is the only recent version that has good port coverage.  (Of a total
> of 1045 ports, 1042 support perl5.16, 137 perl5.18, 10 perl5.20).  This
> amounts to current version, last version and most recent EOL version.
> 
>  * revert as soon as possible Mojca's recent commit that broke perl5.20
> builds until the there has been more discussion and the
> details/consequences worked out.  This is no comment about the merit of
> his suggestion, just that it breaks things.  If he would set up a
> separate test branch, I would be happy to help test.  This would allow
> progress on adding perl5.20 ports.
> 
>  * work toward filling out perl5.18 and perl5.20 ports with the goal of
> moving to a more recent default version as soon as coverage permits.  If
> you're going to add a perl5.18 subport, add the perl5.20 as well so it
> doesn't have to be done twice.
> 
> Just some broad strokes but it would be nice to come to an agreement
> that would allow us to move forward.  Comments?

I think we have historically resisted changes here because we're afraid of breaking things - but in effect have ended up with things that are broken beyond our ability to fix them without causing disruption (due to lack of interest in doing what amounts to really tedious work).

One reason why we have so many more p5.16 ports is that p5.16 was the default perl (it's also worth noting that the perl5 portgroup used to itself list which perl5 versions would be set for each p5 port - but that was switched to require the p5 ports to opt in).

I would guess that a large number of p5 ports probably build fine with perl5.20 - but just haven't had someone go in and specifically edit them (in fact, this would be something relatively easy to test).

We don't have a 'real' need for older perls, as far as I can see.

(we can suffer some pain now, and find/fix broken modules or we can continue to have our confusing mis-mash of unsupported perls and be stuck several versions behind current...)

--
Daniel J. Luke                                                                   
+========================================================+                        
| *---------------- dluke at geeklair.net ----------------* |                          
| *-------------- http://www.geeklair.net -------------* |                          
+========================================================+                        
|   Opinions expressed are mine and do not necessarily   |                          
|          reflect the opinions of my employer.          |                          
+========================================================+