On 05/27/2013 04:15 AM, Marc Lehmann wrote:
On Mon, May 27, 2013 at 11:11:49AM +0200, Michael Panteleit
<[email protected]> wrote:
There is ample evidence against that with 5.18. I suspect your definition
of "healthy and alive, however, includes "breaking compatibility and
support for existing perl code in favour of experimental toys", in which
you would be right of course.
You are singling out a single problem here, ignoring the bulk of them.
The change in hash randomisation alone broke ~100 modules for example,
and the trend didn't start with 5.18, 5.10 already started introducing
widespread XS breakage that to my knowledge has not been fixed either. As
I have a lot of modules and am writing modules for decades, I am very
sensitive to such changes, even if they are mostly invisible to you - it's
me who is doing the work.
The only difference between 5.18 and earlier is the massive amount of
incompatibilities caused by it. The accelerated release schedule also
means that a lot of very badly thought-out changes to the language have
been introduced, canged, changed again etc. (given/when being lexical
while other block constructs are dynamic, the fiasco with smart matches
and many more). It also means a lot of extra work for cpan authors who
have to work around breakage in perl every few months now, which causes a
great additional burden.
That might well be what you mean with "keeping it alive", but ignoring the
enourmous bulk of code that exists and *works*, and relies on a certain
amount of stability, is what keeps perl alive for many other people,
including me.
Changing perl into a python or other language clone doesn't seem to help
it (thats my personal opinion of course). And of course, python doesn't
introduce so widespread breakage so regularly either.
And this section you quoted is quite telling - there is no grammar for
perl, except in the source code. The documentation only works by giving a
few examples, not a formal grammar.
There are a lot of features in perl that are not documented, but are
absolutely vital for it to function (for example ::-quoting wasn't (or
isn't) documented, or the function argument evaluation order, or a large
part of the XS interface - in the past, when I reported such bugs, they
were considered bugs. Apparently that changes now).
If perl5porters now think that perl coders have to read the perl source
code for an idea on whats legal and what not, and at the same time
think that breaking anything that isn't documented according to their
interpretation of the day, then perl isn't healthy or alive, it becomes
useless for productive work.
And yes, that is a change. When I last submitted a patch to fix unicode
bugs in perl I had to make a survey of the whole of cpan to see what
modules break. The standards have changed dramatically since then.
All this aside, the fact remains that the Perl modules for urxvt are
broken for Arch Linux users as well as anyone else running Perl 5.18.
There are modules that address the problem regarding the hash ordering
changes and the other fixes should be trivial. I can empathize with your
frustration, I have 760 Perl modules in my care on the Arch Linux AUR, I
had to rebuild each one, twice, once to get rid of the 5.16 != 5.18 API
mismatches between the XS changes, and once to simply increment the
PKGBUILDs to trigger an 'upgrade' for all the users, and I still have a
trove of 81 modules that either require fixing by hand or upstream
attention. Am I upset? meh, no it's not really worth it, I work with one
of the gentlemen on perl5-porters and his defense of the actions taken
with regard, make it hard to dispute the pros and cons... AND the
perldeltas have in fact been warning about all these changes for quite
some time... That said, I feel that not addressing the problem and
simply getting upset over the changes does us users no good. When can we
expect these issues to be resolved? Urxvt is *the* terminal for ppl who
want the most out of their terminal experience.. none of hte rest even
come close to meeting my needs the way that urxvt does.
Please note that if I missed the fix I apologize, but the machine that
has been upgraded to 5.18 has a very neutered urxvt installation which I
would love to get reverse Mastectomized ASAP. Please advise.
--
John D Jones III
Perl/Javascript/Systemd Zealot
[email protected]
http://www.zoelife4u.org/
_______________________________________________
rxvt-unicode mailing list
[email protected]
http://lists.schmorp.de/cgi-bin/mailman/listinfo/rxvt-unicode
