Geth nqp: a5619aa1a0 | (Christian Bartolomäus)++ | src/vm/jvm/runtime/org/perl6/nqp/runtime/Ops.java
[JVM] Return VMNull from context introspection ops

This avoids explosions in S02-names/pseudo-6e.t (on the JVM backend) that seem to be related to the code in 'method ctx()' in src/core.e/PseudoStash.pm6. (Compare comments in the closed PR
  github.com/rakudo/rakudo/pull/3130.)
06:58
nine bartolin_: looks like the VMNull object unlocked quite a few fixes on JVM :) 08:47
bartolin_: I got the chance to talk to jnthn a bit in Erlangen. Looks like in the long term we'll want to move large parts of spesh up into NQP. This would make it realistic for JVM to get spesh support, too. 08:49
bartolin_: the part that will still have to be handled will probably be the collecting of stats which sounds possible to get into the JVM implementation as well. 08:50
Same is obviously true for the JS backend
[Tux] Rakudo version 2020.02.1-205-g78371bb0c - MoarVM version 2020.02.1-47-g3c3ad0678
csv-ip5xs0.708 - 0.711
csv-ip5xs-206.305 - 6.318
csv-parser25.664 - 26.256
csv-test-xs-200.373 - 0.380
test7.651 - 7.862
test-t2.014 - 2.099
test-t --race0.805 - 1.004
test-t-2033.119 - 33.486
test-t-20 --race9.639 - 10.115
09:27
lizmat Files=1305, Tests=111216, 206 wallclock secs (28.25 usr 8.25 sys + 2894.68 cusr 271.49 csys = 3202.67 CPU) 09:38
Geth rakudo: ea4bdf947a | (Elizabeth Mattijsen)++ | 4 files
Move R:I:CHECKSUM back to IO::Path

But mark it as an implementation-detail, so that we're clear on its status.
10:50
AlexDaniel weekly: perl6→raku progress: #perl6-dev now also redirects to #raku-dev (finally! #perl6→#raku redirection was set in place months ago) 13:37
notable6 AlexDaniel, Noted! (weekly)
Geth ¦ roast: AlexDaniel unassigned from mryan Issue Investigate absence of tests for :dba adverb github.com/Raku/roast/issues/238 14:01
nqp: AlexDaniel unassigned from pmichaud Issue Do we need both cclass 'n' and 'nl'? github.com/Raku/nqp/issues/82
AlexDaniel unassigned from pmichaud Issue Make JSON parsing faster github.com/Raku/nqp/issues/101

  AlexDaniel unassigned from pmurias Issue If `token ws` defined by user, we should treat it as declarative github.com/Raku/nqp/issues/443
9266fb5378 | (Daniel Green)++ | src/NQP/Optimizer.nqp Set return type of intifying a list or hash
This optimization was missed when changing the default NQP behavior from numifying things to intifying them.
AlexDaniel weekly: perl6→raku progress: nqp, nqp-configure and roast repos were moved: github.com/Raku/nqp github.com/Raku/nqp-configure github.com/Raku/roast 14:06
notable6 AlexDaniel, Noted! (weekly)
AlexDaniel weekly: redirects are in place so there doesn't seem to be a need to do anything, but please change links where possible 14:11
notable6 AlexDaniel, Noted! (weekly)
MasterDuke hm, i don't have permission to push to Raku/nqp? 14:17
AlexDaniel: ping 14:21
AlexDaniel pong
MasterDuke do i need to do something to get permission to Raku/nqp? 14:22
AlexDaniel imgur.com/a/RMuHX7i
#softwaregore
MasterDuke did you do something? or did it just take a little while for permissions/accesses to sync? because ^^^ was another attempt at a push, not a delayed notification of the first 14:25
AlexDaniel no, I did something
we should review the permissions at some point
currently it defaults to `Read` for everyone which is probably a good default, but then for every repo we have to explicitly give Write to Raku team
but that's probably a bit too permissive, I'm not sure 14:26
not worse than not being able to contribute because you don't have the rights, so it's still good
travis-ci NQP build passed. Daniel Green 'Set return type of intifying a list or hash 14:40
travis-ci.com/Raku/nqp/builds/154097749 github.com/Raku/nqp/compare/a5619a...66fb537852
Geth rakudo: 9429728cbd | (Elizabeth Mattijsen)++ | 11 files
Remove unnecessary implementation detail

  &SLICE_HUH somehow lived in the SETTING:: but didn't need to.
Move it to Rakudo::Internals and adapt the code and the keys tests accordingly.
15:06
Geth nqp: 019e162fad | (Daniel Green)++ | src/NQP/Optimizer.nqp
When we're (int|num)ifying a list or hash...

re-write into an nqp::elems(). Spesh can do this too, but if we can do it earlier, why not do so.
15:37
MasterDuke jnthn: ^^^ anyway to know if the QAST::Var is actually a list or hash if the sigil is '$'? 15:38
jnthn Implement a type inferencer? :) 15:45
MasterDuke i've got a spare 5-10 min, sure, why not
MasterDuke and if it turns out to take longer than that, instead i guess i'll change all the +$ in nqp to nqp::elems (where appropriate) 15:47
jnthn: oh, and why isn't it safe to unconditionally set the return type of (int|num)ify to int|num? i.e., why is that only done in the case of lists and hashes? 15:49
jnthn MasterDuke: Not sure, maybe it is? 15:52
MasterDuke heh, i'll run a spectest and see what happens 15:53
Geth nqp: bb00e60e67 | (Daniel Green)++ | src/NQP/Optimizer.nqp
Some Optimizer changes

Move all the op type annotation enrichments together.
Set num annotation for '_n' ops even if they're not a num_to_int candidate.
... (5 more lines)
16:15
travis-ci Rakudo build errored. Elizabeth Mattijsen 'Remove unnecessary implementation detail 16:34
travis-ci.org/rakudo/rakudo/builds/664433437 github.com/rakudo/rakudo/compare/e...29728cbd89
lizmat restarted failing job 16:49
tellable6 2020-03-19T13:20:45Z #moarvm <brrt> lizmat what would be vrurg's last work
MasterDuke jnthn: is there any benefit to using $ instead of @ for a list in nqp? e.g., github.com/Raku/nqp/blob/master/sr...FA.nqp#L53 17:06
jnthn Don't think so 17:10
MasterDuke so unless there's a fix for spesh, would you prefer changing $'s to @'s (e.g., $!states to @!states) or changing +$ to nqp::elems()? 17:12
jnthn Maybe the @ communicates better that it's an array 17:16
MasterDuke that was my preference also 17:17
bartolin_ nine: (plan to move large parts of spesh into nqp) that sounds interesting. thanks for the info! 18:44
Geth nqp: usev6++ created pull request #606:
[JVM] Always provide file and line in backtrace
20:23
bartolin_ ^^ spectest is still running, but maybe someone could check if I understood and "translated" this code from MoarVM correctly? github.com/MoarVM/MoarVM/blob/3c3a...#L555-L572 20:27
("file" in annotations defaults to empty string, "line" defaults to 1) 20:28
Geth nqp: e4767444b2 | (Christian Bartolomäus)++ | src/vm/jvm/runtime/org/perl6/nqp/runtime/Ops.java
[JVM] Fix indentation
21:01
Geth nqp: 05339d1b52 | Coke++ | 3 files
Track (some) Perl 6 -> Raku changes
22:07
nqp: 182fb9ed93 | Coke++ | t/docs/opcodes.t
Track changes to opcode doc file

Avoids over 100+ false negatives
22:12
nqp: e2d8be3b0b | Coke++ | t/docs/opcodes.t
find more opcodes in moarvm

This finds some opcodes that were marked as not in moar but then also finds more opcodes that now need documentation
22:46
Kaiepi i may have found a way to optimize a bunch of small spots in rakudo 23:06
if tests pass, IO::Handle.lines will be 10% faster, for instance
MasterDuke nice 23:11