Front page | perl.perl5.porters |
Postings from September 2010
On Sep 5, 2010, at 1:01 PM, Father Chrysostomos wrote: > This is a regression introduced in the past week by yours truly, with change 10cea945821. > > $ perl5.13.4 -le '$a = *foo; open $fh, ">", \$a; print $fh, "b"; print $a' > b > $ bleadperl -le '$a = *foo; open $fh, ">", \$a; print $fh, "b"; print $a' > *main::foo > $ ls > GLOB(0x8017b8) > > bleadperl behaves the same way a 5.8.x. It seems it was changed unintentionally in 5.10.0. In any case, the 5.10/12 behaviour is less surprising and more useful. > > Note that I am only talking about copies of globs here, that is, globs with the FAKE flag on. (That was why I said ‘real globs’ in the commit message for 10cea945821, but I forgot regenerate the patch after updating my working copy.) Here is a patch to fix this. Note that this also affects \*$glob_copy, which is controversial, but at least I am only restoring the 5.10/12 behaviour in this case. (I think ideally that an explicit * should do a SvFAKE_off temporarily [effectively], but in this case the result has no practical application, so it’s not worth the time.)Thread Next