github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
00:14 Altai-man joined 00:16 sena_kun left 00:51 Kaiepi joined 02:15 sena_kun joined 02:17 Altai-man left 03:02 MasterDuke left 03:08 patrickb left 03:17 patrickb joined 03:24 guifa2 left 04:14 Altai-man joined 04:17 sena_kun left 04:42 Prince213 left 04:44 Prince213 joined 05:32 Prince213 left 05:42 patrickb left
nwc10 good *, #moarvm 06:40
07:08 brrt joined
brrt \o nwc10 07:09
nwc10 o/
I am failing to replicate [Coke]'s failures, summarised here: gist.github.com/coke/353adbf0ab607...b4847e312f 07:10
07:15 sena_kun joined 07:16 Altai-man left 07:30 leont joined 07:56 zakharyas joined
nwc10 Oh the awesome. This is building with clang on a Raspberry Pi: 08:17
ccache clang -c -o dyncall_call.o dyncall_call.S
./dyncall_call_arm32_arm_armhf.S:62:2: error: invalid instruction, did you mean: fldmiax? fldmiad r3, {d0-d7} ^
oh, thanks irssi
anyway, build fail due to an illegal instruction
whereas gcc works
[Coke]: what clang version does your machine have? :-)
and did you run the NQP tests? :-) 08:18
evalable6: say 0୯ 08:23
evalable6 Potential difficulties:
Leading 0 has no me…
nwc10, Full output: gist.github.com/f583499a01f9d7e4a4...39ae2419b8 08:24
nwc10 "Use of Nil in numeric context" 08:40
[Coke]: do t/nqp/031-grammar.t and t/nqp/106-unicodenames.t fail for you? 08:45
09:05 AlexDaniel` left 09:13 AlexDaniel` joined 09:18 MasterDuke joined 09:26 brrt left 09:31 Kaiepi left 09:32 Kaiepi joined
nwc10 and the culprit is 05a0e836da2f39d5a7b22722b7c0918179921d73 09:42
10:14 Altai-man joined 10:17 sena_kun left
nwc10 - hashtable->key_right_shift = (8 * sizeof(MVMuint64) - initial_size_base2); 10:21
+ hashtable->key_right_shift = (8 * sizeof(MVMuint32) - initial_size_base2);
10:50 brrt joined 11:08 MasterDuke left 11:11 Kaiepi left 11:12 Kaiepi joined 11:32 zakharyas left
Geth MoarVM/ZZZ-A-better-hash: 52 commits pushed by (Nicholas Clark)++
review: github.com/MoarVM/MoarVM/compare/3...8c289f5601
11:51
nwc10 jnthn: this pig is almost airworthy
[Coke]: that should fix the bugs you found 11:52
dogbert17: that should avoid the aliasing warning. Which happened to appear on a line that was not changed
it's rebased onto split-flags so
1) it should avoid that race
2) it needs the split-flags branch (ie commit) for rakudo - see github.com/MoarVM/MoarVM/pull/1336 11:53
jnthn: in particular, could you look at the commit that starts XXX, as that has some possivly "brave" assumptions.
and now I am required to be "AFK" 11:58
although this time, probably "we" have not just SMS, but even data!
(until this evening)
12:05 MasterDuke joined 12:15 sena_kun joined 12:17 Altai-man left 13:15 Kaeipi joined 13:16 Kaiepi left 14:10 MasterDuke left 14:14 Altai-man joined 14:17 sena_kun left 14:33 dogbert17 left 14:48 brrt left 14:56 zakharyas joined 15:01 zakharyas left 15:02 zakharyas joined
nwc10 and back. Did I miss anything? :-) 15:45
timotimo lightning storm in here 15:47
jnthn Ooh, we had our one of those earlier in the week. It was very loud.
timotimo this was not very loud, but very rainy and very windy 15:48
jnthn nwc10: I didn't get to look over the commit yet; lots of $dayjob bits
nwc10 :-(
jnthn Though I've got tomorrow set aside for Raku/MoarVM things
nwc10 ah OK cool.
although @other can probably look at #1336 and see if *it* is sane. (er, possibly one $other) 15:49
timotimo should i checkout and build zzz and test it for stuff? 15:51
nwc10 timotimo: if you could, that would be wonderful. It needs the raudo branch (as detailed in PR #1336) 15:52
timotimo yeah, the one that switches the flags into two bits
nwc10 in particular, could you check that I didn't somehow screw up the heap profiler or the debug server?
timotimo i can try! 15:53
nwc10 they compile (ship it!) but I don't know if
a) they still work
b) they actually worked on master before I started :-/
what comes after Z? :-) 15:55
timotimo AAAAAA-Better-Hash 15:56
nwc10 evalable6: $a = "ZZZ"; ++$a; say $a
evalable6 (exit code 1) 04===SORRY!04=== Error while compiling /tmp/P36Eo93dp0
Variable '$a' is not declared
at /tmp/P36Eo93dp0:1
------> 03<BOL>08⏏04$a = "ZZZ"; ++$a; say $a
nwc10 evalable6: my $a = "ZZZ"; ++$a; say $a
evalable6 AAAA
timotimo m: say "ZZZ".succ
camelia AAAA
nwc10 AAAAhaa
odd, I seem to have lost a commit that added a hash size to MVM_str_hash_build() 16:04
timotimo what does the reflog say?
nwc10 I've found it in XXX-A-Better-Hash 16:05
I've sort of lost count of how many checkout I have across how many machines.
timotimo forgot to put "install" after make 16:12
16:15 sena_kun joined 16:17 Altai-man left 16:36 brrt joined 16:57 brrt left 17:15 zakharyas left
Geth MoarVM/ZZZ-A-better-hash: 7c0f3ddf51 | (Nicholas Clark)++ | 8 files
MVM_str_hash_build now takes an entry count for the hash.
17:45
nwc10 that's the "missing" commit. 17:46
timotimo it's too warm 17:57
[Coke] nwc10: didn't run nqp tests, went straight to rakudo. one moment 18:11
nwc10: Apple clang version 11.0.3 (clang-1103.0.32.62)
... switching to ZZZ, trying again.
18:14 Altai-man joined
[Coke] nqp tests all pass in zzz 18:16
rakudo build fails, however.
18:16 sena_kun left
[Coke] gist.github.com/coke/89ff8592af70c...e74ac55c67 18:17
dies in the c code for nqp ops.
looks like we need to track the flags -> flags1/2 change. 18:18
I'm using rakudo:master, maybe I need a branch? 18:19
(also using nqp:master)
18:20 AlexDaniel` left 18:27 AlexDaniel` joined
timotimo there's a branch in nwc's personal clone 18:32
but it's just a tiny patch
[Coke] is it all "flags1"?
timotimo +#ifdef MVM_COLLECTABLE_FLAGS1 18:33
+#define RAKUDO_FIRST_FLAG 128
+#else
#define RAKUDO_FIRST_FLAG 16384
+#endif
this as well
[Coke] added his fork as a remote, building... 18:37
(git++)
nwc10 [Coke]: the newest clang I could find was clang version 7.0.1-8 (tags/RELEASE_701/final) 18:41
earlier clang doesnb't optimise hard enough to expose my bug
[Coke] ok. the current build with ZZZ/nqp/your fork, all tests pass in nqp/rakudo! 18:52
nwc10 \o/
ship it! oh, wait, still some XXXes to review
I'm very glad that you exposed that bug.
thanks for testing (twice) 18:53
[Coke] literally the least I could do. :) 19:00
(i mean, and still do something)
nwc10 true, *but* you were diligent at repeating that it was still broken. and you happen to have a platform (that I don't) that tests both a different OS and a compiler that I forgot about. 19:01
[Coke] (y)
(which is apparently MS/Skype shorthand for "thumbs up"
👍 19:02
so I'll just do that, then. :)
19:14 MasterDuke joined
MasterDuke nwc10: moarvm/nqp/rakudo all build and pass tests on ZZZ-A-better-hash. using linux 5.7.10-arch1-1 x86_64, gcc (GCC) 10.1.0 and clang version 10.0.1 19:54
nwc10 cool, thanks 19:57
20:15 Kaeipi left 20:16 Kaiepi joined 20:42 squashable6 left 20:44 squashable6 joined 21:15 sena_kun joined 21:16 Altai-man left 21:18 Kaiepi left 21:20 Kaiepi joined 23:04 leont left 23:06 Kaiepi left 23:09 Kaiepi joined