Re: Optionally / safely disabling a warning category for a complete package (magic numbers)

Sorry for not testing this (which would require multiple versions of Perl that I didn't already have handy). I suspect this will work based on prior similar code I've used:

BEGIN {
    warnings->unimport('experimental::signatures')
        if  eval "no warnings 'experimental::signatures'; 1";
}
[download]

The advantage to this version is that it doesn't presume that the Perl version is a reliable indicator of whether or not some type of warning can be disabled. It is at least possible that some future version of something would mean that an upper bound on the version would need to be added. But I'm also not convinced that you can't have cases where unusual requirements lead to a mismatch between the Perl version and the behavior of "no warnings 'experimental::signatures';". For example, in building Perl for weird environments (like tiny Unix devices) I've ended up with versions of Perl where standard features got disabled because porting of base Perl got finished but not porting of some standard extensions for that version of Perl.

- tye

Comment on Re: Optionally / safely disabling a warning category for a complete package (magic numbers) Download Code

Replies are listed 'Best First'.
Re^2: Optionally / safely disabling a warning category for a complete package (silent failure) by tye (Sage) on May 18, 2016 at 19:39 UTC
Though, this approach also has the disadvantage of silently failing if your "no warnings" should be supported but something weird is just broken. You could mitigate that problem by emitting a warning (or fatal error) if the failure reason doesn't match the one expected, but that can lead to noise when your expectations become incorrect, of course. `BEGIN { my $class = 'experimental::signatures'; if( eval "no warnings '$class'; 1" ) { warnings->unimport($class); } elsif( $@ !~ /Unknown warnings category / ) { warn "'no warnings' failed: $@\n"; # Or even die(). } }` [download] - tye	[reply] [d/l]
Re^2: Optionally / safely disabling a warning category for a complete package (magic numbers) by haukex (Archbishop) on May 18, 2016 at 19:53 UTC
Hi tye, You're absolutely right that magic numbers are not good, that is a potential weakness of my proposed solutions. FWIW, when the postderef feature was taken out of experimental status, it was elected to leave the "experimental::postderef" warnings category in place but to make it a no-op. References: commit, P5P Regards, -- Hauke D	[reply]
Re^3: Optionally / safely disabling a warning category for a complete package (backward compat) by tye (Sage) on May 18, 2016 at 20:37 UTC
FWIW, when the postderef feature was taken out of experimental status, it was elected to leave the "experimental::postderef" warnings category in place but to make it a no-op. As is good practice and as I would expect. But the next thing I would expect to eventually happen would be to clean out really old cruft by making references to 'experimental::postderef' deprecated and cause a warning and then to later make it fatal (by removing all mention of it from the source code). Not that I'll make any predictions about approximate time frames for such. :) - tye	[reply]