On Sun May 30 05:44:14 2010, masak wrote:
> <masak> I've found a case where a line number is one-off. should I
> nopaste it?
> <jnthn> masak: please
> <masak> I don't think it's the fault of the new backtraces thing, to
> be honest.
> <masak> more like a result of parsing or something.
> <jnthn> Well, the other thing is that we don't have as many :node($/)s
> in Actions.pm as we maybe should.
> <jnthn> So could very well be some wins to be had by some careful
> extra scattering of those.
> <masak> http://gist.github.com/418990
>
> (inlined here)
>
> $ cat one-off
> sub bar {
> pfff
> }
>
> bar
>
> $ perl6 one-off
> Could not find sub &pfff
> in 'bar' at line 1:one-off
> in main program body at line 6:one-off
>
> <jnthn> masak: curious
> <jnthn> masak: Does it get it right if you do pfff() ?
> <jnthn> as in, add parens?
> * masak tries
> * jnthn had a test very much like that which got it right
> <masak> no.
> <jnthn> ok
> <masak> maybe it sees the statement as starting on line 1, or
> something?
> <jnthn> That's very possible.
> <jnthn> I just thought that it'd been getting it right.
> <jnthn> (I had a very similar test case...)
> * jnthn plays spot the difference
> <jnthn> masak: hmm, I agree it seems to be one off there.
> * masak submits rakudobug
> <jnthn> masak: Think it may be a Parrot fix.
> <jnthn> masak: Thing is, the annotations in the PIR look correct.
> <jnthn> Trying my first guess at a patch now.
> <masak> I'll include that in the Rakudobug. :)
We needed a Parrot tweak which I did a couple of days back - but that
only got us off by one in the *other* direction because it emitted the
right annotation, but we were emitting the wrong line number in it.
Grrr. Anyway, today I nailed it with a little tweak in Rakudo, so now:
C:\Consulting\rakudo>type test.p6
sub bar {
pfff
}
bar
C:\Consulting\rakudo>perl6 test.p6
Could not find sub &pfff
in 'bar' at line 2:test.p6
in main program body at line 5:test.p6
Which is spot on. \o/
Given to moritz++ for testing.
Thanks,
Jonathan
