00:09 Altai-man joined 00:11 sena_kun left 00:14 Altai-man left 00:50 frost-lab joined 01:07 MasterDuke left 01:48 frost-lab left
vrurg_ .ask Altai-man do you know why blin always dies with "terminated by signal SIGKILL (Forced quit)"? Each time it's a little above 1000 passed tests. 01:50
tellable6 vrurg_, I'll pass your message to Altai-man
timotimo vrurg_: check dmesg for the oom killer? 02:13
02:14 vrurg_ is now known as vrurg
vrurg timotimo: that's it. Thanks! Now I need to find out what's this. :) 02:15
timotimo not good :(
vrurg Ok, the VM needs more memory. 02:16
Not bad either. It's a blin-dedicated disposable VM.
timotimo there's stuff moarvm could do "better" in cases where parts of working memory tend to be swapped out 02:17
though now that i think about it, it's probably not easy at all 02:18
vrurg Things like "MoarVM panic: Internal error: invalid thread ID 193 in GC work pass" worry me more for now. 02:21
timotimo yes, that's more not good
vrurg As I currently running a LOT of tests in parallel, I see moar panicing one way or another too many.
But more fun are recent errors of this kind: "No such method 'append' for invocant of type 'Array[Str:D]'. Did you mean 'append'?" 02:22
timotimo uh oh 02:24
vrurg Overall, as I spent last two weeks fixing concurrency in Test::Async and Async::Workers, the picture is rather depressing. I saw ThreadPoolScheduler re-invoking continuations, realloc issue, SIGSEGV, broken react/whenever... 02:25
timotimo realloc issue and sigsegv can both be a sign of general memory corruption 02:26
so out-of-bounds accesses and concurrent access to non-concurrency-safe structures
i assume you're on 2020.10 versions? 02:28
vrurg Sure. I'm not doing C for too long now, but I'm a concurrency fun since writing some threaded utils for OS/2. :)
Right, 2020.10
02:35 go|dfish left 02:37 lucasb left 02:40 go|dfish joined
vrurg 5 rakudo/moar failures over 20000 runs. Neither terrible nor good. 02:42
timotimo do you happen to have core dumps? though they are usually too late to actually be helpful 02:45
vrurg timotimo: unfortunately, no. I'm forgetting to allow core dumps. 03:53
timotimo: pardon me, I was looking in a wrong place. I do have like 4.1 gigs of those. 03:55
06:02 greppable6 left, unicodable6 left, quotable6 left, linkable6 left, coverable6 left, squashable6 left, nativecallable6 left, bloatable6 left, tellable6 left, benchable6 left, bisectable6 left, committable6 left, sourceable6 left, releasable6 left, notable6 left, evalable6 left, shareable6 left, statisfiable6 left 06:03 statisfiable6 joined, coverable6 joined, squashable6 joined, evalable6 joined, notable6 joined 06:04 unicodable6 joined, releasable6 joined, greppable6 joined, bloatable6 joined, sourceable6 joined, linkable6 joined, benchable6 joined 06:05 tellable6 joined, quotable6 joined, bisectable6 joined, shareable6 joined, nativecallable6 joined, committable6 joined 06:20 raku-bridge1 joined, raku-bridge1 left, raku-bridge1 joined 06:21 raku-bridge left, raku-bridge1 is now known as raku-bridge 07:52 Altai-man joined 07:56 patrickb joined 08:09 sena_kun joined 08:11 Altai-man left 08:20 domidumont joined 08:52 jmerelo joined
jmerelo I just created the rakulang group in GitLab, after Gitlab donated a "gold" subscription to us. In principle, the intention was to use it to hold pipelines, but of course we can use it for anything. 08:52
(raku was already taken by a dude since 2017) 08:53
Please dm or email me your Gitlab account so that I can add you with the appropriate permissions.
I'm sending invitations to a few of you I know are there (jjatria, tyil) also. 08:55
tyil jmerelo: ack :> 08:56
jmerelo: is there any particular reason or goal with the gitlab namespace (as compared to the github one that's already in use) 08:58
jmerelo tyil: I can bump you up to owner if you want, since you are more familiar with it.
08:59 nebuchadnezzar left
tyil I'm not against it, but I need to know what to do with it ;~; 08:59
08:59 nebuchadnezzar joined
jmerelo tyil: as said above, we entered a mini-panic when we realized we were over the free tier in Travis, which we use heavily. So we tried to get CI resources elsewhere. In principle, it's mainly for that kind of thing. 08:59
tyil ah ok 09:00
I know some gitlab CI if you want help with that, and I heavily use gitlab registries at work for docker too
jmerelo We can think about creating mirrors there if needed. But there are other things that are possible: adopting modules that are hosted in GitLab (not a lot are, but there are a few) if it's required by the owner, for instance. 09:01
tyil: I didn't even know there were gitlab docker registries...
tyil jmerelo: mirroring the module list for zef might be nice, so people can add their modules either on github or gitlab (if thats not the case yet) 09:02
jmerelo BTW, welcome to the RSC, tyil++
tyil: definitely _not_ the case. It should be a two-way mirror, then.
tyil: it would probably be so much easier to just have two different lists... 09:03
tyil hmm yeah, that's probably true 09:06
also thanks :>
09:10 lizmat left, lizmat joined
lizmat Files=1337, Tests=114215, 225 wallclock secs (29.46 usr 8.62 sys + 3098.12 cusr 301.85 csys = 3438.05 CPU) 10:34
10:45 frost-lab joined 11:20 Kaiepi left
lizmat bisectable6 11:41
bisectable6: role A[*@types] does Positional {method new(*@a) { self.CREATE }; method STORE(*@a) { } }; my @b is A[Int,Int] = 42,666; dd @b
bisectable6 lizmat, Will bisect the whole range automagically because no endpoints were provided, hang tight
lizmat, Output on all releases: gist.github.com/2cd4b0f3591917f551...f9f9e58378
lizmat, Bisecting by exit code (old=2020.05.1 new=2020.06). Old exit code: 1
lizmat, bisect log: gist.github.com/caa13fd91be47ede1c...d2276ad1c9 11:42
lizmat, (2020-05-12) github.com/rakudo/rakudo/commit/de...5ff3690813
lizmat, Bisecting by output (old=2018.12 new=2019.03.1) because on both starting points the exit code is 1
lizmat, bisect log: gist.github.com/b0d6c86f917201c886...e6d8dcd13e
lizmat, Bisecting by output (old=2018.06 new=2018.08) because on both starting points the exit code is 1
lizmat, bisect log: gist.github.com/e9ea84a5dc7721ec2a...ca18aa363e 11:43
lizmat, (2018-07-19) github.com/rakudo/rakudo/commit/e9...c636fb166a
lizmat, Output on all releases and bisected commits: gist.github.com/9afd1a40f0c6f321db...607e26cc19
nine releasable6: status 11:48
releasable6 nine, Next release in ≈15 days and ≈7 hours. 1 blocker. Changelog for this release was not started yet
nine, Details: gist.github.com/802a476b08fe5ead03...9605f7cb1f
nine I'm gonna merge github.com/MoarVM/MoarVM/pull/1373 to give it enough time before the release 11:49
lizmat nine: will you bump NQP / Rakudo as well? 12:07
or should I?
nine feel free :) 12:08
12:09 Altai-man joined
lizmat oki 12:11
12:11 frost-lab left 12:12 sena_kun left
Geth_ nqp: 217d541512 | (Elizabeth Mattijsen)++ | tools/templates/MOAR_REVISION
Bump MoarVM to get nine++ internment fixes
Altai-man nine, the branch can be deleted? 12:18
tellable6 2020-11-06T01:50:37Z #raku-dev <vrurg_> Altai-man do you know why blin always dies with "terminated by signal SIGKILL (Forced quit)"? Each time it's a little above 1000 passed tests.
nine guess so
Altai-man done 12:19
Geth_ rakudo: 5552434572 | (Elizabeth Mattijsen)++ | tools/templates/NQP_REVISION
Bump NQP to get nine++ internment fixes
13:59 patrickb left
jmerelo The Raku group in gitlab is already provisioned with the Gold subscription. Anyone wants to be added to the group, just let me know. 14:03
14:03 jmerelo left
[Coke] .tell jmerelo - I am @coke on gitlab 14:13
tellable6 [Coke], I'll pass your message to jmerelo
[Coke] Do we need to setup gitlab repos that mirror the github ones? 14:50
ugexe why was there a mini panic over travis? the entire reason i added circle-ci support was because back then i said travis was going downhill and we should just stop using it. combined with patrickbs CI work I dont really see an issue with travis 14:53
and you can add a module from gitlab/farthub whatever to the regular ecosystem list. META.list is just a list of links to META6.json, nothing else 14:56
if instead its some sort of stallman-esq requirement to host the projects.json on gitlab then any user can already modify the zef config to point to some other gitlab instance 14:58
imo i dont think we really gain a significant number of developers by complicating or changing our toolset/workflow compared to the number we potentially lose due to increased complexity and lower recognition from recruiters (they dont care if you prefer github, they know github and thats what they will check) 15:03
[Coke] I thought the github work was to get access to other CI tooling if we wanted it, not to move the module list, but it's hard to get full context from irc. 15:05
sigh. the "gitlab" work.
ugexe we have azure CI and OBS for CI (as well as circle ci which will probably go away) 15:06
we don't really need a 6th CI
[Coke] fair enough. 15:07
Would probably help if we had all the various extant options documented. 15:08
regarding Azure - is it through an org?
(and if so, do we have some kind of extra credit as an oss project?)
ugexe i believe so. patrickb is basically the master CI orchestrator right now 15:09
[Coke] .ask patrickb can you clue me in to the current Azure setup we have (and/or permit me to the subscription?) 15:14
tellable6 [Coke], I'll pass your message to patrickb
ugexe github.com/rakudo/rakudo/pull/4000...1339370387 15:16
example azure results in github
dev.azure.com/Rakudo/rakudo/_build...ew=results (and one of the examples directly on azure) 15:17
github.com/rakudo/rakudo/blob/mast...elines.yml -- the CI file... although this one is not very basic since it tests lots of variations like relocating rakudo etc 15:21
[Tux] Rakudo v2020.10-53-g555243457 (v6.d) on MoarVM 2020.10-31-gb13f6a24c
csv-ip5xs0.825 - 0.836
csv-ip5xs-207.813 - 8.004
csv-parser25.241 - 25.327
csv-test-xs-200.383 - 0.387
test7.591 - 7.652
test-t1.785 - 1.815
test-t --race0.820 - 0.932
test-t-2030.625 - 30.684
test-t-20 --race9.653 - 9.717
16:00 Kaiepi joined 16:09 sena_kun joined 16:11 Altai-man left 16:18 domidumont left 17:08 domidumont joined 19:00 domidumont left 19:13 vrurg left 19:36 vrurg joined 20:05 MasterDuke joined
MasterDuke j: say "works" 20:05
camelia works
MasterDuke j: my $a = 42; my $b = 42; $a does role { method Str() { "foo" } }; $b.Str.say 20:06
camelia foo
MasterDuke j: my $a = 1; my int $b = 1; $a does role { method Str() { "foo" } }; $b.Str.say
camelia 1
MasterDuke m: my $a = 42; my $b = 42; $a does role { method Str() { "foo" } }; $b.Str.say 20:07
camelia foo
MasterDuke m: my $a = 1; my int $b = 1; $a does role { method Str() { "foo" } }; $b.Str.say
camelia foo
MasterDuke we should probably decide which is correct and write some tests
timotimo yeah, we have a cache for small integers
[Coke] cache or no cache, I think leaking that role is wrong 20:08
tellable6 2020-11-06T17:58:24Z #raku <jmerelo> [Coke] on its way
timotimo there could probably be a double-check when getting an object out of that cache if the type actually matches
if it doesn't match, nuke the cache entry
20:09 Altai-man joined
[Coke] I assume the cache is required speed wise? (have we tested that recently) 20:09
timotimo shouldn't be difficult to take it out, there's a few ops that spesh emits that directly go to the cache, though 20:10
nwc10 I've been having a look already - hence the question 20:11
but not "take it out" - see if there's a better plan
20:11 sena_kun left
MasterDuke yeah, there was a small, but noticeable slower and increase in mem used if the cache was disabled 20:12
timotimo oh, only a small slowdown? 20:13
what was the test workload like?
MasterDuke nine and i were chatting about this a day or two ago. i think 0.5s slower to parse CORE.c and 20mb more ram used 20:14
starting here colabti.org/irclogger/irclogger_lo...-11-03#l36 20:15
timotimo i somehow completely missed that 20:19
20:21 MasterDuke left 20:37 MasterDuke joined 20:51 patrickb joined
patrickb ugexe, [Coke]: Azure provides its service for free for OSS projects without any limits. The current Azure CI setup tests all build combinations that Travis, CircleCI and AppVeyor tested. So from a coverage point of view Azure replaces all the others. The others are only still active because we didn't disable them yet and to gain some experience with Azure. 20:57
tellable6 2020-11-06T15:14:04Z #raku-dev <[Coke]> patrickb can you clue me in to the current Azure setup we have (and/or permit me to the subscription?)
2020-11-06T16:03:58Z #raku <melezhik> patrickb: Config::Parser::toml test OK on Rakudist - rakudist.raku.org/sparky/report/debian/912 , which version of Rakudo do you use?
timotimo patrickb: you don't happen to have a setup to test raku builds on mingw? well, just moarvm should be enough
patrickb timotimo: I don't. But I think it should be doable. If you want you can give it a try. Azure has a nice web editor for it's configuration file. :-) 20:59
[Coke]: We currently have three separate organizations registered on Azure: MoarVM, Raku and Rakudo (named identical to their GitHub counterpart). 21:02
[Coke]: The owner of each of these orgs is [email@hidden.address] that's an account currently maintained by rba. 21:03
[Coke]: The interesting part of the configuration is done by editing the azure-pipelines.yml present in each of the rakudo, nqp and MoarVM repo though. 21:04
[Coke] Do you have an Azure account? I can invite you in each of the orgs. 21:06
ugexe, [Coke]: I detailed my plan for our CI in github.com/Raku/problem-solving/pull/219 In short: I'm actively working on a tool to act as an intermediary between GitHub and Azure + OBS. This will make the OBS test results visible and hopefully allow us to implement a reasonable, reliable and usable CI workflow. 21:13
MasterDuke patrickb: is it possible to restart individual azure jobs? sometimes there's a fail because of a network timeout or something like that and it'd be nice to be able to restart, but i don't see that option now 21:15
patrickb melezhik: I think tyil already has a lead on what might have been wrong. In fact I managed to test the module successfully as well after manually installing its dependencies.
tellable6 patrickb, I'll pass your message to melezhik
[Coke] patrickb: [email@hidden.address]
patrickb MasterDuke: I think there is. I'm just unsure if one needs to be logged in to azure to have access to the button. 21:16
[Coke] patrickb: do you find we have enough azure cycles to get things done?
patrickb (The tool I'm working on will automate restarting jobs, so this will be a non-problem then.)
MasterDuke cool 21:17
patrickb [Coke]: Not sure I understand the question. Azure does not impose any limits. Neither on number of parallel jobs, nor on maximum hours per month.
[Coke] Sorry, used to the sub model where you pay for everything. 21:19
patrickb [Coke] I invited you to all three orgs. 21:20
[Coke] patrickb: thanks.
patrickb Have to leave. Please do ask questions wrt our CI and the plans for it. I'm very willing to explain.
(and I do backlog)
\o 21:21
21:21 patrickb left
[Coke] patrickb: clicked on the "join now" in each email, in each case, taken to a page that gives me a 401 not allowed. (all mentioning my [email@hidden.address] account) 21:21
tellable6 [Coke], I'll pass your message to patrickb
[Coke] o/
signed out and back in, same error. 21:23
401 - Uh-oh, you do not have access.
21:35 squashable6 left 21:38 squashable6 joined 21:50 squashable6 left
ugexe MasterDuke: you can restart jobs within the github interface for azure jobs (like the little icons that show red or green for prs) but the button does not show up unless its a failed build 21:50
21:51 squashable6 joined
MasterDuke hm, i've looked around but never seen a way. let me try again 21:58
ugexe: so right on this page i should be able to github.com/MoarVM/MoarVM/pull/1344 ? 22:00
the red x's aren't clickable for me 22:01
ugexe click on 'Details'
(the red x is on the left, 'Details' is on the right)
i dont have write perms to moarvm though so i don't see the button on these specifically 22:02
MasterDuke oh, now i see the 're-run'. thanks 22:03
22:11 MasterDuke left 22:24 finsternis joined 23:33 Altai-man left