Front page | perl.perl5.porters |
Postings from November 2010
2010/10/26 Dave Mitchell <davem@iabyn.com>:
> On Tue, Oct 05, 2010 at 11:52:20PM -0700, Andreas J. Koenig via RT wrote:
>> b2a2a9010bb3413ad9c32e455d93e01069d0fd73
>> Author: David Mitchell <davem@iabyn.com>
>> Date: Mon Oct 4 15:18:44 2010 +0100
>>
>> stop map,grep leaking temps [perl #48004]
>>
>> breaks the tests of SARTAK/MooseX-Role-Parameterized-0.19.tar.gz with a
>> bunch of error diagnostics, starting with
>
> I *think* this is due to a bug (or rather, unreasonable expectations)
> in that distribution's test suite. In particular, it has code roughly
> like:
>
> sub foo {
> map { SomeClass->new->some_method($_) } @_
> }
>
> i.e. it creates a temporary object each time through the map block, then
> discards it. The change in blead means that the temporary object is freed
> at the end of each map iteration, rather than at the next statement
> boundary following the map. This means that the object's destructor is
> called earlier, which makes the tests fail.
>
> The only question really is whether this is an acceptable change in blead.
> I personally think it is. Opinions, anyone?
We never guaranteed when destruction happens.
But I dislike the early temporary object destruction as it looks like much
slower inside map, to afterwards. Do we really need that?
--
Reini Urban
Thread Previous
|
Thread Next