Front page | perl.perl5.porters |
Postings from May 2015
On Thu, Apr 30, 2015 at 05:23:55PM -0600, Karl Williamson wrote:
> On 04/30/2015 02:17 PM, (Andreas J. Koenig) (via RT) wrote:
> ># New Ticket Created by (Andreas J. Koenig)
> ># Please include the string: [perl #124430]
> ># in the subject line of all future correspondence about this issue.
> ># <URL: https://rt.perl.org/Ticket/Display.html?id=124430 >
> >
> >
> >Another one from Slaven:
> >
> >bisect
> >------
> >commit 273e254d1663b9223905e4b5c3b6546671ba365e
> >Author: Karl Williamson <khw@cpan.org>
> >Date: Tue Mar 10 13:16:23 2015 -0600
> >
> > Optimize out unicode_to_native(), native_to_unicode()
> >
> >rt.cpan ticket
> >--------------
> >https://rt.cpan.org/Public/Bug/Display.html?id=103976
> >
> >
>
> I was able to reproduce this problem, and chatted with Zefram on irc about
> it. He believes that the problem lies with
> https://github.com/clkao/invoker. To summarize,
>
> "the necessary and sufficient fix is for invoker_ck_entersub to check that
> it's got an entersub op, and skip most of its check-time logic when it
> doesn't"
>
> Also, "invoker ought to behave as a lexically-scoped pragma, rather than
> globally affecting any use of $-"
>
> And, "this is partly down to the means by which invoker hooks the op
> checker, which is via another CPAN module, the combination of the two is
> broken, and will fail when it meets any call checker that rewrites an
> entersub into some other kind of op. your commit merely introduces another
> such rewriting checker"
>
> I've created
> https://rt.cpan.org/Public/Bug/Display.html?id=104135
> against invoker
BTW, I'm not convinced that the call checker is the right way to optimise
away those functions, in that it is a compile-time check on the CV in the
GV, whereas function calls are normally resolved by a run-time lookup of
the CV in the GV; so for example the following doesn't print "hi":
*utf8::unicode_to_native = sub { print "hi!\n" };
utf8::unicode_to_native($x);
It's probably fairly unlikely that someone will want to override those
functions, and it will still override if the override is done in a
BEGIN/use, but its still not technically correct.
--
If life gives you lemons, you'll probably develop a citric acid allergy.
Thread Previous
|
Thread Next