On Fri Jul 23 10:10:57 2010, coke wrote: > On Fri Oct 09 08:01:32 2009, masak wrote: > > <masak> rakudo: / <[z ......... booga booga - ]> / > > <p6eval> rakudo daf221: OUTPUT«src/gc/api.c:253: failed assertion > > 'PObj_is_PMC_TEST(obj)'[...] > > <masak> \o/ > > <pmichaud> rakudo: / <[a-z]> / > > <p6eval> rakudo daf221: OUTPUT«src/gc/api.c:253: failed assertion > > 'PObj_is_PMC_TEST(obj)'[...] > > <pmichaud> that looks to me like a p6eval issue > > <pmichaud> or an exceptions problem > > <moritz_> $ >perl6 -e '/ <[a-z]> /' > > <moritz_> Attributes of type 'struct llsig_element **' cannot be > > subclassed from a high-level PMC. > > <moritz_> still LTA > > <pmichaud> $ ./perl6 -e '/ <[a-z]> /' > > <pmichaud> perl6regex parse error: Unescaped '-' in charlist > (use '..' > > or '\-') at offset 5, found '-' > > <moritz_> pmichaud: are you on latest master? > > <pmichaud> no. > > <moritz_> since llsig is a new concept, it might well make a > difference > > <jnthn> What on earth is it trying to do with a sig there. :-/ > > <moritz_> anyway, we need masak for that one ;-) > > <masak> what? what am I supposed to do? :P > > <moritz_> bug it! > > <jnthn> .oO( oh BUGger ) > > <pmichaud> the "Unescaped '-' in charlist" is the correct output. > > This is now slightly more awesome in Atlanta, but doesn't quite match > STD. > > 13:10 <@[Coke]> std: / <[a-z]> / > 13:10 <+p6eval> std 31794: OUTPUT«===SORRY!===Unsupported use of - as > character range; in Perl 6 please use .. > at /tmp/lqLPfKUdfV > line 1:------> / <[a-z⏏]> /Parse failedFAILED 00:01 > 116m» > 13:10 <@moritz_> rakudo: /<[a-b]>/ > 13:10 <+p6eval> rakudo b2af27: OUTPUT«===SORRY!===Obsolete use of > hyphen in > enumerated character class;in Perl 6 please use .. > instead at > line 22, near "b]>/"» >
Fixed. Turns out we had it very close in rakudo, differing by a single space in the error message. But this error was coming from nqp-rx (fixed), which then had to be re-imported into parrot (done), then rakudo needed to be updated to use the later version of parrot (done), and a test in roast to make sure we didn't lose it (done.) I don't particularly like the direct test against the compile-time error text, but given the often stated desire that we want the error messages to be consistent with STD... -- Will "Coke" Coleda
