Front page | perl.perl5.porters |
Postings from July 2010
??var Arnfj??r?? Bjarmason wrote:
>As of f746176000 (Bareword sub lookups) by Zefram there's a regression
>in how perl handles signals:
Turns out this commit is not actually introducing a new bug. It's making
an old bug happen in more circumstances than previously. Here's a
shorter test for the new situation:
$ perl5.13.3 -lwe '$SIG{ALRM} = sub { warn "OK\n" }; eval "sub x {Y::SIG->z};1"; alarm 1; sleep 2'
Signal SIGALRM received, but no signal handler set.
$ perl5.10.1 -lwe '$SIG{ALRM} = sub { warn "OK\n" }; eval "sub x {Y::SIG->z};1"; alarm 1; sleep 2'
OK
And here's the equivalent test for a closely related situation:
$ perl5.13.3 -lwe '$SIG{ALRM} = sub { warn "OK\n" }; eval "sub x {\@Y::SIG->z};1"; alarm 1; sleep 2'
Signal SIGALRM received, but no signal handler set.
$ perl5.10.1 -lwe '$SIG{ALRM} = sub { warn "OK\n" }; eval "sub x {\@Y::SIG->z};1"; alarm 1; sleep 2'
Signal SIGALRM received, but no signal handler set.
$ perl5.004 -lwe '$SIG{ALRM} = sub { warn "OK\n" }; eval "sub x {\@Y::SIG->z};1"; alarm 1; sleep 2'
Signal SIGALRM received, but no signal handler set.
-zefram
Thread Previous