[00:42] *** timo left [00:55] *** kjp left [01:08] *** kjp joined [06:20] *** kjp left [06:20] *** apogee_ntv left [06:20] *** apogee_ntv joined [06:21] *** kjp joined [06:24] *** kjp left [06:24] *** harrow left [06:24] *** ShimmerFairy left [06:25] *** ShimmerFairy joined [06:25] *** kjp joined [06:29] *** harrow joined [14:54] *** [Coke] left [16:51] *** [Coke] joined [18:33] *** MasterDuke joined [18:37] timo: still around? reading the logs? i don't understand this data i'm logging. i'm printing the name that's being looked up in MVM_frame_find_lexical_by_name, and a rakudo backtrace. in the two examples i've run with this logging turned on (my log parsing code and building part of rakudo), '$?CLASS' is by far the most common. but most of the [18:37] MasterDuke, I'll pass your message to timo [18:37] methods that are at the top the backtrace don't have any reference to it at all! [18:38] 2070 $?CLASS - !cursor_capture [18:38] 2980 $?CLASS - !cursor_fail [18:38] 4158 $?CLASS - !cursor_start_cur [18:38] 60746 $?CLASS - find_method [18:38] 5209060 $?CLASS - !cursor_start [18:38] ^^^ from building part of rakudo [18:38] oops, no, that's from my log parsing example [18:38] 181164 $?CLASS - type_check [18:38] 194585 $?CLASS - [18:38] 464446 $?CLASS - list [18:38] 553309 $?CLASS - visit_children [18:38] 590099 $?CLASS - EXPR [18:39] ^^^ *that's* from building part of rakudo [18:40] !cursor_start used to be 11m before my PR was merged. but i don't know why it isn't even less now... [18:44] there are no explicit references to $?CLASS in any EXPR, or visit_children, or a bunch of the others [18:49] i wondered if it was inlining, but i tried with MVM_SPESH_INLINE_DISABLE=1 and also with --optimize=0 (for rakudo), but no difference [18:59] Something checking that a method is being invoked on an expected class? Perhaps as part of a method cache? [18:59] *** MasterDuke left [19:48] .tell MasterDuke did you see https://github.com/Raku/nqp/commit/e5e5a905f1 ? [19:48] lizmat, I'll pass your message to MasterDuke