Front page | perl.perl5.porters |
Postings from November 2022
On Thu, 10 Nov 2022, 15:14 Felipe Gasper, <felipe@felipegasper.com> wrote:
> Hi all,
>
> Sometime back I merged a change to make `sub {}` parse to
> OP_ANONCODE rather than OP_SREFGEN/OP_ANONCODE. There’s been some CPAN
> breakage from this change, documented here:
> https://github.com/Perl/perl5/issues/20384
>
> Paul graciously fixed some of the breakage. I have a PR pending
> that I think addresses the rest: https://github.com/Perl/perl5/pull/20462
>
> That PR, however, will still leave at least 2 CPAN modules broken
> because, as a side-effect of my original change, OP_DESC for a coderef will
> now be “anonymous subroutine” rather than “single ref constructor”. For
> example:
>
> -----
> > perl5.34 -w -e 'sub {}'
> Useless use of single ref constructor in void context at -e line 1.
>
> > bleadperl -w -e 'sub {}'
> Useless use of anonymous subroutine in void context at -e line 1.
> -----
>
> In my opinion the latter is an improvement; however, those CPAN
> modules--I think there are 2--that expect the former in their tests would
> need updates.
>
> Is this an acceptable behaviour change?
>
IMO Yes. Normally we try hard to provide fixup patches for things like
this, but it is inevitable fallout of improving diagnostics.
Yves
>
Thread Previous