Front page | perl.perl5.porters |
Postings from March 2013
Jesse Luehrs wrote: >This seems unintentional. Indeed. The Perl_ck_warner() call in die_unwind() used to happen before unwinding, so would be affected by the lexical warning state at the die() site. Now it happens after unwinding, so takes the lexical warning state at the catching site. I don't have a clear idea of which behaviour is more correct. t/op/die_keeperr.t, which was introduced as part of my exception handling changes, is actually testing for the catching-site criterion, but that's not asserting that the criterion should be that. The documentation speaks of "no warnings 'misc'", but doesn't say which lexical scope matters. Assuming we want to revert this change, the easy fix is to move the conditional Perl_ck_warner() back to before unwinding. A more difficult way would be to determine the disposition of the warning before unwinding and then warn in the required manner after unwinding. I see no compelling reason to warn after unwinding rather than before, so just moving the warning code should be fine. Patch attached. -zeframThread Previous | Thread Next