> On Dec 10, 2021, at 12:19, Dave Mitchell <davem@iabyn.com> wrote: > > On Thu, Dec 09, 2021 at 04:58:57PM -0500, Dan Book wrote: >>> Having read the prior arguments, I'm kinda with Dave on this one. >>> @DB::args is already shockingly broken in a lot of ways. Making it >>> always appear empty when called within a signatured sub is, honestly, >>> not the worst of its already-current breakages. >> >> This isn't going to be much consolation to users who are currently relying >> on it and have not experienced the bugs. > > > But @DB::args (and thus croak() stack trace argument display etc) would > only disappear on signature subs. No existing code not using experimental > features would be affected. > > And it's only ever been intended for use in debugging etc, not in > production. Debugging is the context where it _matters_, though, to have as much detail as possible. Having the arguments in stack traces has been a boon over the years. We know the liabilities this will entail … do we yet know any more about the potential performance wins from it? If they’re significant, then it’s at least a fair trade: trickier debugging for the sake of a palpable speed bump. But if the speed benefits aren’t going to make a real-world difference, then maybe leave @_ as it is? At the very least, args-less Perl stack traces will be no _worse_ than the args-less stack traces in other languages I’ve debugged. -FGThread Previous | Thread Next