Front page | perl.perl5.porters |
Postings from April 2013
On Sun, Mar 10, 2013 at 12:48:16AM +0000, Dave Mitchell wrote:
> However, I would expect something like the following *should* work and be
> safe, but it doesn't:
>
> use re 'eval'; return qr/(?{ $RGP::Success = 1 })$$re/;
>
> And as you pointed out, a 'use re eval' even in the scope of the main
> qr/foo/ doesn't turn off the fatal error. So there's definitely something
> there needing more work.
The branch smoke-me/davem/re_overload, currently being smoked and shortly
to be merged into blead, should fix this.
If the following change is made to your test script:
--- /tmp/rgtest.pl.orig 2013-04-11 10:28:53.879421869 +0100
+++ /tmp/rgtest.pl 2013-04-11 10:31:57.510778815 +0100
@@ -18,7 +18,7 @@
package RGP;
use overload q{""} => sub {
my $re = shift;
- return '(?{ $RGP::Success = 1 })' . $$re;
+ return qr/(?{ $RGP::Success = 1 })$$re/;
}, fallback => 1;
}
then it passes both tests when using my new branch. To make it still work
on 5.16.x and earlier, you also need a use re eval:
- return '(?{ $RGP::Success = 1 })' . $$re;
+ use re 'eval';
+ return qr/(?{ $RGP::Success = 1 })$$re/;
which is harmless, but unnecessary after my fix to blead.
I haven't tried fixing up Regex::Grammars to see if it works, since it's
too complex for a cursory glance and fix-up. I see lots of places in the
code where it returns strings containing '(?{', but I don't know how
practical it would be to convert all those to return qr//'s instead.
--
Gravity is just a theory; teach Intelligent Falling in our schools!
http://www.theonion.com/content/node/39512
Thread Previous
|
Thread Next