[perl #74414] [BUG] Multiple dispatch doesn't treat 'rw' parameters as narrower than readonly parameters in Rakudo

[jn...@jnthn.net via RT]

On Tue Nov 10 07:55:06 2015, pesc...@gmail.com wrote:
> On Sat Feb 22 17:37:47 2014, coke wrote:
> > On Fri Apr 16 01:31:22 2010, masak wrote:
> > > <masak> rakudo: multi foo(Int $x is rw) { say "writable" }; multi
> > > foo(Int $x) { "readonly" }; foo(42)
> > > <p6eval> rakudo 78faa0: OUTPUT«Ambiguous dispatch to multi 'foo'.
> > > Ambiguous candidates had signatures:␤:(Int $x)␤:(Int $x) [...]
> > > * masak submits rakudobug
> > > <masak> S06: "To allow modification, use the C<is rw> trait.  This
> > > requires a mutable object or container as an argument (or some kind
> > > of
> > > type object that can be converted to a mutable object, such as
> > > might
> > > be returned by an array or hash that knows how to autovivify new
> > > elements). Otherwise the signature fails to bind, and this
> > > candidate
> > > routine cannot be considered for servicing this particular call.
> > > (Other multi candidates, if any, may succeed if they don't require
> > > C<rw> for this parameter.)"
> > > <moritz_> now you're just showing off :-)
> > > <masak> :P
> >
> > Behavior is mostly unchanged since 2010 - error message has better
> > signature reporting, but still reported as ambiguous.
> >
> > 20:36 <[Coke]> m: multi foo(Int $x is rw) { say "writable" }; multi
> > foo(Int $x)
> >                { "readonly" }; foo(42)
> > 20:36 <camelia> rakudo-moar b2b333: OUTPUT«Ambiguous call to 'foo';
> > these
> >                 signatures all match:␤:(Int $x is rw)␤:(Int $x)␤  in
> > sub foo at
> >                 /tmp/CwM3ZkL1nu:1␤  in block  at /tmp/CwM3ZkL1nu:1␤␤»
> 
> With Rakudo commits 8cba8863 and ed10656c this now behaves as follow:
> 
> 15:52 < psch> m: multi foo(Int $x is rw) { say "writable" }; multi
> foo(Int $x) { "readonly" }; say foo(42); foo my $ = 42;
> 15:52 <+camelia> rakudo-moar ed1065: OUTPUT«readonly␤writable␤»
> 
> I've also added S06-multi/by-trait.t to spectest.data which tests
> this.

psch++ also updated most of the multi-dispatch caching logic to get things 
performing well again. Today I updated the dynamic optimizer in Moar to 
understand the new multi-dispatch flag, so we don't go mis-inlining. And fixed 
a further static optimizer issue where it wasn't aware of `is rw` and so could 
mis-dispatch. Both covered by more tests in S06-multi/by-trait.t.

So, we can consider this done.

/jnthn