Perl 6 - the future is here, just unevenly distributed

IRC log for #moarvm, 2017-11-18

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
01:30 bisectable6 joined #moarvm
01:51 Ven joined #moarvm
02:24 timotimo https://lemire.me/blog/2017/11/16/fast-exact-integer-divisions-using-floating-point-operations/  and  https://github.com/ridiculousfish/libdivide
02:24 timotimo i'm not sure how often we have a division by a known integer number, but maybe we can teach the jit a few division fastpaths?
02:25 timotimo well, step one is surely to turn some divisions into bit shifts
02:56 ilbot3 joined #moarvm
02:56 Topic for #moarvm is now https://github.com/moarvm/moarvm | IRC logs at  http://irclog.perlgeek.de/moarvm/today
08:21 domidumont joined #moarvm
08:26 domidumont joined #moarvm
08:56 robertle joined #moarvm
10:15 evalable6 joined #moarvm
10:15 nativecallable6 joined #moarvm
10:15 quotable6 joined #moarvm
10:15 tangible6 joined #moarvm
10:15 releasable6 joined #moarvm
10:15 unicodable6 joined #moarvm
10:15 bloatable6 joined #moarvm
10:15 benchable6 joined #moarvm
10:15 committable6 joined #moarvm
10:15 greppable6 joined #moarvm
10:15 coverable6 joined #moarvm
10:15 squashable6 joined #moarvm
10:15 bisectable6 joined #moarvm
10:15 statisfiable6 joined #moarvm
10:20 lizmat joined #moarvm
11:02 Ven joined #moarvm
11:04 lizmat joined #moarvm
12:07 nine Yet another observation: it's probably not a terribly good idea to delete the only instruction of a basic block. This may be a reason why keeping a no_op instead helps (though not the only one).
12:40 nebuchadnezzar joined #moarvm
12:42 lizmat joined #moarvm
13:22 dogbert17 I guess that tests which fail when using MVM_SPESH_NODELAY=1 MVM_SPESH_BLOCKING=1 can't be considered as blockers
13:22 jnthn No, if they were we'd not have had a release for a couple of years. :P There's a *lot* less of them now than there were before, though.
13:23 jnthn And most, last I checked, are thanks to stack traces losing frames when inlines happen
13:23 dogbert17 indeed, it's hard work finding them :P
13:23 jnthn That's fixable, just a bit tedious
13:24 dogbert17 in this case it was in t/spec/S09-typed-arrays/native-num.t
13:24 dogbert17 test #247, 'Cannot resolve caller is-approx(Str, Str); none of these signatures match:'
13:26 dogbert17 btw, what about brrt's fix from yesterday, will that be included in the release?
13:27 jnthn We can cherry-pick it I guess
13:27 jnthn We should not merge that branch though
13:27 jnthn I don't want my other commit in the releas
13:27 jnthn *release
13:27 jnthn Hm, I wonder if it's on me to do the release this time :)
13:30 lizmat joined #moarvm
13:39 dogbert17 at least there aren't any real blockers this month
13:40 dogbert17 unless AlexDaniel manages to dig one up at the last second
13:57 MasterDuke joined #moarvm
15:11 AlexDaniel yeah, unfortunately toaster didn't manage to finish while I was sleeping, so had to kick it off again
15:12 AlexDaniel but it should be ready soon. Once I see the results I think we'll be good to go
15:15 AlexDaniel jnthn: I was assuming that it's not, but at the same time I haven't heard from samcv yet
15:19 timotimo AlexDaniel: any +/- on the commit i sneaked in last night?
15:19 timotimo or you jnthn? as this is basically your code?
15:19 AlexDaniel timotimo: I've looked at it, but it's hard for me to predict what problems this could possibly cause
15:20 timotimo if any user code decides to return () from the additional-names method, it'll have exploded before and will explode even with the patch attached
15:20 timotimo applied*
15:20 AlexDaniel greppable6: alternative-names
15:20 greppable6 AlexDaniel, https://gist.github.com/d47791c04f07a7a629ff46830cd0e43e
15:21 timotimo it'd be cool to have a little summary to go along with the link, like "two lines matched" and "from only one module: doc"
15:22 AlexDaniel https://github.com/perl6/whateverable/issues/259
15:22 AlexDaniel unicodable6 already does something similar, should be very easy to implement
15:23 AlexDaniel unicodable6: infinity
15:23 unicodable6 AlexDaniel, U+1AB2 COMBINING INFINITY [Mn] ( ᪲)
15:23 unicodable6 AlexDaniel, U+221E INFINITY [Sm] (∞)
15:23 unicodable6 AlexDaniel, 5 characters in total: https://gist.github.com/f074033df8c67bd23a8a0f926cb17928
15:23 AlexDaniel well… it doesn't show all of the characters if there are some weird ones :)
15:23 timotimo fair enough
15:25 AlexDaniel timotimo: I think your commit is OK. Thanks for notifying
15:26 timotimo oh, a quick coverage run tells me the register function is not run during actual startup
15:35 zakharyas joined #moarvm
15:39 zakharyas joined #moarvm
16:01 MasterDuke joined #moarvm
16:08 zakharyas1 joined #moarvm
18:11 AlexDaniel speaking of blockers :) https://github.com/rakudo/rakudo/issues/1254
18:11 AlexDaniel .tell samcv you're probably interested in https://github.com/rakudo/rakudo/issues/1254
18:11 yoleaux AlexDaniel: I'll pass your message to samcv.
18:11 domidumont joined #moarvm
18:12 AlexDaniel I guess this is the offending commit: https://github.com/MoarVM/MoarVM/commit/e76788819
18:31 jnthn Ah, there be blockers
18:32 AlexDaniel unfortunately I had a busy week and couldn't get my hands on it earlier, sorry for that :)
18:33 jnthn No worries, I was just thinking "oh man, I should make dinner but I should maybe work on cutting the release". Now the latter ain't an option anyway. ;)
18:44 jnthn With luck, samcv++ might have chance to comment/look at that. I fnot I guess we can always revert it, but I dunno if other stuff depends on it...
18:45 jnthn *If not
18:45 jnthn I'll have a bit of time for release tonight, or could do it tomorrow also
18:49 AlexDaniel I think tomorrow is a more realistic bet
18:50 ugexe i'm not sure what will happen with https://github.com/libuv/libuv/issues/1625 - since node already made a release with the regression they might not revert
18:51 AlexDaniel ugexe: oh… I have not seen that
18:55 AlexDaniel ugexe: so, does it mean that we should revert 8b53b1e for now?
18:55 AlexDaniel Well, that's exactly what you said, but do you still think so? :)
18:56 evalable6 joined #moarvm
19:08 AlexDaniel fwiw I think it's much better to file tickets like this: https://github.com/rakudo/rakudo/issues/1255
19:09 AlexDaniel (instead of hoping that I'll find a comment in a pull request that points to a bug report in a different repo, for an issue that we probably want to handle before the release)
19:33 samcv good * everyone
19:33 yoleaux 18:11Z <AlexDaniel> samcv: you're probably interested in https://github.com/rakudo/rakudo/issues/1254
19:34 samcv that is a weird issue
19:34 samcv time to add another roast test
19:35 samcv AlexDaniel: it's still passing roast right
19:35 AlexDaniel yeah
19:35 samcv never seen this before
19:35 AlexDaniel do I agree that it is weird, the case seems so down to earth that it's really surprising there was no test or something
19:36 AlexDaniel and only *one* ecosystem module is failing because of this
19:36 samcv Cn is unassigned. and i fixed unassigned characters but i would have think that it would have been caught hm. maybe we need tests for every single general category
19:36 samcv i just woke up. will look at it in 30 mins or so
19:37 AlexDaniel samcv: it's alright, take your time
19:37 AlexDaniel samcv: by the way, is it you making the release this time or jnthn?
19:37 samcv i can do the MoarVM release again yeah
19:38 AlexDaniel (some context: https://irclog.perlgeek.de/moarvm/2017-11-18#i_15466864 )
19:40 AlexDaniel jnthn: objections? :)
19:47 samcv AlexDaniel: the reason i think ascii letters must come up Cn is that they have no bitfield row
19:48 samcv but not 100% sure yet
19:51 samcv oh looks like non ascii also show Lu. argh
19:53 AlexDaniel yeah I didn't test much
19:54 AlexDaniel samcv: fwiw Unicode::PRECIS is the module that has failing tests because of this
19:56 samcv this is really weird
19:56 samcv ok so Cn is the default value as it should be. (index 0) and Ll is index 1
19:56 samcv but lowercase letters work but not uppercase
19:57 samcv oh wow. cause "Lu" is totally missing… that's not good
20:02 AlexDaniel u: X
20:02 unicodable6 AlexDaniel, U+0021 EXCLAMATION MARK [Po] (!)
20:02 unicodable6 AlexDaniel, U+0036 DIGIT SIX [Nd] (6)
20:02 AlexDaniel waaat…
20:02 unicodable6 AlexDaniel, 1733 characters in total: https://gist.github.com/6dbffcb19ef6975b15345e9ded3afa36
20:03 AlexDaniel u: aX
20:03 domidumont joined #moarvm
20:03 unicodable6 AlexDaniel, U+2F44 KANGXI RADICAL AXE [So] (⽄)
20:03 unicodable6 AlexDaniel, U+A009 YI SYLLABLE AX [Lo] (ꀉ)
20:03 unicodable6 AlexDaniel, 56 characters in total: https://gist.github.com/fe5b9e2c80410459104467328b71ca66
20:03 AlexDaniel u: ∞X
20:03 unicodable6 AlexDaniel, U+221E INFINITY [Sm] (∞)
20:03 unicodable6 AlexDaniel, U+0058 LATIN CAPITAL LETTER X [Lu] (X)
20:03 AlexDaniel I guess unicodable is using some older version
20:03 AlexDaniel unicodable6: uptime
20:03 unicodable6 AlexDaniel, 9 hours, 48 minutes, and 44 seconds, 343.48KB maxrss. This is Rakudo version 2017.10-8-g6ac53e427 built on MoarVM version 2017.10 implementing Perl 6.c.
20:03 AlexDaniel yeah
20:27 zakharyas joined #moarvm
20:29 jnthn samcv++
20:30 jnthn samcv: If you're up for doing the release, that'd be really nice :)
20:31 AlexDaniel jnthn: what are your thoughts on https://github.com/rakudo/rakudo/issues/1255 ?
20:32 jnthn iiuc, we already shipped 2017.10 with that bump in place
20:32 AlexDaniel oh… right.
20:32 * AlexDaniel completely missed that
20:33 samcv well i think i found what's happening. well not exactly but sort of
20:33 jnthn oh, wait, that may not be the csae
20:33 jnthn *case
20:33 jnthn https://github.com/MoarVM/MoarVM/pull/719
20:33 jnthn It was filed longer ago, then merged 21 days ago, so it wasn't in 2017.10
20:33 samcv i see  $points_by_hex->{0041} ( for 'A') being set and can print it out. but then lower down it is undefined and i'm not sure why
20:34 AlexDaniel I see, that's an old commit but merged recently.
20:34 AlexDaniel jnthn: so it's probably reasonable to revert it for now?
20:34 jnthn Was filed on 8th Oct, on 10th Oct I said OK to merge after upcoming release, which woulda been 2017.10
20:35 jnthn I think so, yes. This effectively means we are blocked from any libuv bumps until this issues is fixed upstream.
20:35 AlexDaniel or we find a workaround, but that's ok
20:35 jnthn Or that, yeah
20:36 jnthn But given there's a good chance it impacts more than just us...
20:36 * jnthn checks the libuv ticket
20:36 AlexDaniel nobody complained so far
20:36 AlexDaniel at least according to that ticket
20:36 japhb AlexDaniel: I think unicodable6 is off in its units for memory use.  I suspect it's supposed to be MB, not KB.
20:36 japhb (or maybe MiB)
20:37 AlexDaniel japhb: yes, it is. But it is also using nqp::getrusage when it should be using something else (whatever we have now available)
20:37 japhb Telemetry, IIRC
20:38 jnthn I think lizmat discovered that the units getrusage returns aren't consistent across platforms
20:38 jnthn And may well have worked around that in Telemetry
20:38 jnthn AlexDaniel: I'm going to merge the revert from ugexe++
20:38 Geth ¦ MoarVM: 17cbb59876 | (Nick Logan)++ (committed using GitHub Web editor) | 2 files
20:38 Geth ¦ MoarVM: Revert "Use uv_fs_copyfile API"
20:38 Geth ¦ MoarVM: review: https://github.com/MoarVM/MoarVM/commit/17cbb59876
20:38 Geth ¦ MoarVM: 79d6f20057 | (Jonathan Worthington)++ (committed using GitHub Web editor) | 2 files
20:38 Geth ¦ MoarVM: Merge pull request #752 from ugexe/revert-719-uv_fs_copyfile
20:38 Geth ¦ MoarVM:
20:38 Geth ¦ MoarVM: Revert "Use uv_fs_copyfile API"
20:39 Geth ¦ MoarVM: review: https://github.com/MoarVM/MoarVM/commit/79d6f20057
20:39 jnthn That gives some space to figure it out
20:39 AlexDaniel jnthn: good. Thank you
20:40 AlexDaniel japhb: https://github.com/perl6/whateverable/issues/260
20:43 lizmat joined #moarvm
20:44 AlexDaniel ugexe: so we're good now, right?
20:44 AlexDaniel (with that particular proc issue)
20:45 ugexe yes. i also believe ive just proved the regression exists inside nodejs now too
20:47 AlexDaniel ugexe++
20:48 lizmat japhb: what does "say VM.osname" say for you wrt MB vd KB in Telemetry ?
20:54 lizmat perhaps we need more of a heuristic approach to finding out what the unit is
21:02 samcv ok it seems to be assigning Lu and Cn to the same propvalue
21:02 zakharyas joined #moarvm
21:02 samcv getting closer to figuring out how to fix this
21:03 samcv and in doing so decided to nuke the $points_by_hex reference hash. since it's stupid to have $points_by_code AND $points_by_hex
21:23 japhb lizmat: I was noticing unicodable6 doing it in channel, not something on my machine
21:24 lizmat japhb: ?? how ?
21:24 lizmat link?
21:24 japhb Hold on a sec
21:25 japhb https://irclog.perlgeek.de/moarvm/2017-11-18#i_15467144
21:26 lizmat yeah, that looks to be off by a factor of 1024  :-)
21:27 AlexDaniel the code is here: https://github.com/perl6/whateverable/blob/61a995c389a838109c9c7cdda62af717be38b870/lib/Whateverable.pm6#L161
21:27 lizmat unicodable6: help
21:27 unicodable6 lizmat, Just type any unicode character or part of a character name. Alternatively, you can also provide a code snippet. # See wiki for more examples: https://github.com/perl6/whateverable/wiki/Unicodable
21:27 AlexDaniel but it's pre-telemetry code
21:28 lizmat maybe the problem is in the "KB" :-)
21:29 AlexDaniel c: HEAD use Telemetry; say max-rss, ‘ KB’
21:29 committable6 AlexDaniel, ¦HEAD(149ed24): «===SORRY!=== Error while compiling /tmp/w3abLsSZSI␤Undeclared routine:␤    max-rss used at line 1␤ «exit code = 1»»
21:29 AlexDaniel c: HEAD use Telemetry; say T<max-rss>, ‘ KB’
21:29 committable6 AlexDaniel, ¦HEAD(149ed24): «451880 KB»
21:29 lizmat m: use nqp; say nqp::atpos_i(nqp::getrusage,4)   # says 69623808 for me
21:29 camelia rakudo-moar 149ed2429: OUTPUT: «76004␤»
21:30 AlexDaniel c: HEAD use nqp; say nqp::atpos_i(nqp::getrusage,4)
21:30 committable6 AlexDaniel, ¦HEAD(149ed24): «451964»
21:30 AlexDaniel yeah, it's ‘KB’ that is a problem :)
21:31 lizmat m: use Telemetry; say T<max-rss> # says 85640 for me
21:31 camelia rakudo-moar 14a90e044: OUTPUT: «92964␤»
21:33 ugexe node -e "const cp = require('child_process'); { const bat = cp.spawn('cmd.exe', ['/k', 'dir'], { stdio: 'inherit', detached: false, windowsHide: true }); }" # here is the proof of libuv regression inside the latest nodejs release, which went undetected by all of: libuv tests, nodejs test matrix / system, rakudo tests, rakudo spectests.
21:34 AlexDaniel oh… but to get telemetry I'd need to update rakudo
21:35 AlexDaniel so expect a bot kaboom
21:36 mst ugexe: some bugs just hate everybody at once.
21:38 bloatable6 joined #moarvm
21:38 quotable6 joined #moarvm
21:38 tangible6 joined #moarvm
21:38 bisectable6 joined #moarvm
21:38 committable6 joined #moarvm
21:38 nativecallable6 joined #moarvm
21:38 releasable6 joined #moarvm
21:38 coverable6 joined #moarvm
21:38 greppable6 joined #moarvm
21:38 evalable6 joined #moarvm
21:38 benchable6 joined #moarvm
21:38 unicodable6 joined #moarvm
21:38 statisfiable6 joined #moarvm
21:38 squashable6 joined #moarvm
21:38 AlexDaniel unicodable6: uptime
21:38 unicodable6 AlexDaniel, 27 seconds, 213.712MB maxrss. This is Rakudo version 2017.10-214-g14a90e044 built on MoarVM version 2017.10-82-ge876f1484 implementing Perl 6.c.
21:38 japhb lizmat: Ah, so Telemetry auto-corrects the getrusage difference?
21:39 japhb AlexDaniel++
21:39 japhb lizmat++
21:39 mst AlexDaniel: may be worth considering sticking ZNC in front of the bothead code like simcop does with perlbot
21:39 lizmat yes, afaik, only MacOS returns bytes rather than Kbytes
21:40 AlexDaniel mst: yes, absolutely. https://github.com/perl6/whateverable/issues/261
21:41 samcv i think i've almost fixed it
21:41 mst also I strongly suspect he'd be happy to answer questions
21:44 jnthn ugexe++ # bug hunting
21:44 AlexDaniel japhb: https://github.com/perl6/whateverable/commit/e9ccebadca :)
21:46 mst samcv: why are you trying to register perl6 as a group? we did that years ago, just tell me what you need
21:46 AlexDaniel mst: I was also thinking about connecting through Matrix, but the current freenode bridge seems to be a bit slow sometimes
21:46 samcv i didn't do anything
21:46 japhb AlexDaniel: is Telemetry correcting to kB or KiB?
21:46 mst somebody's emailed in
21:47 mst samcv: apparently somebody else attempted to send in a group registration and claimed you were part of the plan
21:47 samcv oh ok
21:48 AlexDaniel japhb: … it should be lowercase k, right?
21:48 samcv well i heard something but i didn't submit anything. someone just told me something about perl 6 masks or something
21:48 AlexDaniel ah, but we don't talk about that anymore.
21:48 samcv but i didn't know what was going on
21:48 AlexDaniel japhb: in any case, docs talk about that: https://docs.perl6.org/type/Telemetry#method_max-rss
21:49 lizmat ah, I completely missed that my Telemetry pod was cleaned up!
21:49 lizmat ++ to whoever did that
21:50 * lizmat will try to update Telemetry doc tomorrow when she's not as tired ad feeling better...
21:53 AlexDaniel japhb: OK. So it is incorrect now, right?
21:54 AlexDaniel japhb: because we take “Kbytes” which I assume is KiB and divide that by 1000…
21:54 * AlexDaniel hates that with passion
21:56 samcv ok going to run spectest one more time then i'll push the Unicode fixes AlexDaniel
21:56 brrt joined #moarvm
22:02 samcv joined #moarvm
22:03 japhb AlexDaniel: Yeah, I would *hope* that getrusage is returning KiB (1024) when it's not returning true bytes as in MacOS ... but certainly taking KiB and dividing by 1000 to get MB is mixing bases.
22:04 Geth ¦ MoarVM: a29bc81b53 | (Samantha McVey)++ | tools/ucd2c.pl
22:04 Geth ¦ MoarVM: unicode: fix uppercase letters not giving Lu for the General_Category
22:04 Geth ¦ MoarVM:
22:04 Geth ¦ MoarVM: ucd2c.pl was trying to set both Cn (Unassigned) and Lu (Letter
22:04 Geth ¦ MoarVM: Uppercase) to the same pvalue code (0). This fixes it so they are not
22:04 Geth ¦ MoarVM: set to the same value.
22:04 Geth ¦ MoarVM: review: https://github.com/MoarVM/MoarVM/commit/a29bc81b53
22:04 Geth ¦ MoarVM: 89fae17a6f | (Samantha McVey)++ | 2 files
22:05 Geth ¦ MoarVM: Commit the updated Unicode database
22:05 Geth ¦ MoarVM:
22:05 samcv \o/
22:05 Geth ¦ MoarVM: Generated with the changes from the previous commit's change to ucd2c.pl
22:05 Geth ¦ MoarVM: review: https://github.com/MoarVM/MoarVM/commit/89fae17a6f
22:05 AlexDaniel samcv: \o/
22:05 AlexDaniel samcv: should I bump moar&nqp?
22:05 samcv held off on some other changes i want to make to ucd2c.pl until after the release
22:05 AlexDaniel or is it coming?
22:05 samcv yes please!
22:05 AlexDaniel ok
22:05 samcv then i'll push an added roast test
22:07 AlexDaniel done
22:07 samcv still amazed this wasn't caught in any of our tests!
22:07 AlexDaniel yes :'(
22:07 samcv i guess nothing tested Lu
22:07 AlexDaniel if not for that module, we'd release it like that :D
22:08 AlexDaniel “maybe”, depending on whether I'd run whateverable tests before the release or not
22:15 samcv i mean really i probably need to take a look at my tests i wrote during my grant and finalize some i didn't end up putting in
22:16 samcv and tweak to make sure property values are reachable
22:17 AlexDaniel I was thinking maybe we should be testing whatever perl6 reports for a character against some database that we know is correct
23:03 nativecallable6 joined #moarvm
23:03 tangible6 joined #moarvm
23:03 committable6 joined #moarvm
23:03 quotable6 joined #moarvm
23:03 greppable6 joined #moarvm
23:03 benchable6 joined #moarvm
23:03 releasable6 joined #moarvm
23:03 evalable6 joined #moarvm
23:03 bloatable6 joined #moarvm
23:03 unicodable6 joined #moarvm
23:03 coverable6 joined #moarvm
23:03 squashable6 joined #moarvm
23:03 bisectable6 joined #moarvm
23:03 statisfiable6 joined #moarvm
23:47 MasterDuke joined #moarvm
23:54 Geth joined #moarvm

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary