sena_kun has broke blin, oh, how wise of you, sena_kun, how wise... 00:20
tellable6 2020-01-08T22:00:06Z #raku-dev <lizmat> sena_kun: also, the default .gist is calling .raku
00:21 lucasb left 00:31 MasterDuke left 01:52 Altai-man_ joined 01:54 sena_kun left 02:08 ufobat__ joined 02:12 ufobat_ left
Xliff .tell vrurg I've got an issue I will need your help with. It looks like Method::Also is now no longer working on method overrides. 03:24
tellable6 Xliff, I'll pass your message to vrurg
vrurg Xliff: ping me tomorrow. 03:25
Xliff vrurg: Will do.
tellable6 2020-01-08T16:51:58Z #raku-dev <patrickb> Xliff Did you try nuking your zef Repo store? I seem to recall this is another instance of the infamous ecosystem-doesn't-support-real-versions problem.
2020-01-08T17:26:32Z #raku-dev <patrickb> Xliff: jnthn++ has just uploaded a new version. When you try again it should just work.
hey Xliff, you have a message:
03:53 sena_kun joined 03:55 Altai-man_ left 04:58 quotable6 left, nativecallable6 left, shareable6 left, evalable6 left, statisfiable6 left, bloatable6 left, notable6 left, committable6 left, reportable6 left, benchable6 left, tellable6 left, releasable6 left, greppable6 left, squashable6 left, sourceable6 left, coverable6 left, bisectable6 left, unicodable6 left 04:59 bisectable6 joined, greppable6 joined, tellable6 joined, bloatable6 joined, committable6 joined, unicodable6 joined, squashable6 joined 05:00 evalable6 joined, sourceable6 joined, quotable6 joined, releasable6 joined, nativecallable6 joined, statisfiable6 joined 05:01 benchable6 joined, shareable6 joined, notable6 joined, coverable6 joined, reportable6 joined 05:52 Altai-man_ joined 05:55 sena_kun left 07:19 ufobat__ is now known as ufobat 07:40 domidumont joined
lizmat :Files=1294, Tests=109644, 211 wallclock secs (28.53 usr 8.46 sys + 2954.98 cusr 278.25 csys = 3270.22 CPU) 07:53
07:53 sena_kun joined 07:55 Altai-man_ left
Geth Blin: 3f3a3e8c99 | Altai-man++ | bin/blin.p6
Fix condition

The smart one-liner Did not consider Mu means `all modules` A silly mistake
09:52 Altai-man_ joined 09:55 sena_kun left
jnthn lizmat: Hm, if I'd had to guess I'd have thought the WhateverCode form would come out the same of potentially even a bit simpler... 10:31
(e.g. simpler generated code so maybe a tad faster)
Can't be much different though; 3% is pretty small :) 10:32
10:46 scovit joined
|Tux| Rakudo version 2019.11-414-g66cf6f811 - MoarVM version 2020.01-15-g1fa9764bf
csv-ip5xs0.725 - 0.742
csv-ip5xs-206.222 - 6.364
csv-parser22.663 - 23.486
csv-test-xs-200.423 - 0.429
test7.362 - 7.386
test-t0 :)
test-t --race0.834 - 0.913
test-t-2029.555 - 30.486
test-t-20 --race9.223 - 9.271
11:00 Guest38485 left 11:07 domidumont left 11:53 sena_kun joined 11:55 Altai-man_ left 11:57 AlexDaniel joined, AlexDaniel left, AlexDaniel joined
lizmat jnthn: yeah, but in a lot of potentially hot code 12:07
but yeah, I would have figured the WhateverCode version would be faster, as it also has fewer allocations 12:08
so it feels to me there is room for (hopefully) simple optimization
nine Then it's probably just some small issue preventing better performance. However "small" doesn't necessarily mean "easy to fix". Could e.g. require PEA to be finished. 12:13
lizmat nine: but what I understand about WhateverCode, is that it can never have any outside references 12:17
or am I wrong there ? 12:18
nine That was just an example for illustrative purposes :) 12:21
12:44 Guest78373 joined 12:46 titsuki left
Xliff nine: Well, initial work on the parallel builder has tentatively yielded a 3-fold decrease in compile times across whole projects, however I still think I'm doing some things wrong. 13:01
Not quite sure what this error means: "P6M Merging GLOBAL symbols failed: duplicate definition of symbol gobject" 13:06
jnthn lizmat: It can close other things, for sure 13:16
13:21 domidumont joined 13:52 Altai-man_ joined 13:54 sena_kun left
AlexDaniel lizmat: what about 14:22
nine: ↑
here's my plan. New members are added to the “Everyone” team, which only has access to this repo
meaning that immediately they're subscribed to it, and receive a notification
lizmat feels like a plan 14:23
AlexDaniel after some time we can auto add them to the Raku team, which will give them access to a lot of stuff
by that time I guess they've had their chance to turn off the feature
lizmat m: (1/10000000000).base(3) 14:24
camelia ( no output )
lizmat m: say (1/10000000000).base(3)
camelia 0.0000000000000000000010
lizmat I think that trailing 0 is incorrect, do we agree? AlexDaniel jnthn nine
AlexDaniel yeah 14:25
c: all say (1/10000000000).base(3)
m: put (1/10000000000).base(3)
camelia 0.0000000000000000000010
committable6 AlexDaniel, ¦all (62 commits): «0.0000000000000000000010␤»
AlexDaniel it's not entirely incorrect but it shouldn't be there
lizmat well, I'm refactoring Rat.base and my solution doesn't have the 0 at the end, but spectest says it should 14:26
but that test also has the comment: # is the trailing zero correct?
AlexDaniel lizmat: interesting. Link to the spectest?
I guess someone was writing the test based on the current behavior
lizmat S32-num/base.t line 44
jnthn Yeah, that trailing 0 feels odd to me also 14:27
lizmat ok, fixing the test :-)
14:31 scovit left
AlexDaniel tyil: I have an idea 14:35
tyil: I'm a bit tired of setting up webhooks for every repo, so, can't we use an org webhook for that?
tyil: however, I think it means that Geth will receive same payloads more than once
tyil: so if we can have a little feature that makes wait a bit before announcing org payloads, it'll be awesome! 14:36
makes it*
I guess the default channel should be #raku 14:37
tyil AlexDaniel: that would be better, I agree 14:39
I currently have a number of other projects I'm spending time on, but I'll add it to my todo list 14:40
idk if I can even differentiate between org and project payloads, btw 14:41
AlexDaniel tyil: we can use a different link 14:42
tyil that should work 14:45
14:48 MasterDuke joined
lizmat m: say (1/10000000000).base(3) # turns out this is actually correct 15:18
camelia 0.0000000000000000000010
lizmat hmmm... or maybe not 15:19
nine AlexDaniel: that's an epic solution! Many thanks! 15:36
Xliff: so you've changed the locking in rakudo? 15:38
Xliff: that message means exactly what it says: we're trying to import a global symbol from a computation unit, but discover that one by that name is already there
lizmat nine: compilation ? 15:39
Guest78373 lizmat: 15:46
lizmat Guest78373: thanks, but it is not about the actual calculation, it's about how many digits should be represented 15:48
Guest78373 aha 15:49
15:55 sena_kun joined, Altai-man_ left
nine lizmat: yeah 16:19
lizmat *phew* thought I had missed a memo somewhere :-) 16:20
16:31 domidumont left 16:49 lucasb joined
AlexDaniel hmm is tony-o around? 16:56
Geth ¦ problem-solving: AlexDaniel assigned to JJ Issue SEO for the keyword "rakulang" seems poor 16:57
17:08 AlexDaniel left
lizmat m: say (1/10000000000).base(3,*) # this shouldn't hang, should it ? 17:21
camelia (timeout)
lucasb ^^ 17:22
lizmat m: say (1/3).base(10,*) # simpler case
camelia (timeout)
lizmat generating 2**63 digits feels meh to me 17:23
I mean, generating more digits than can be represented faithfully in a Rat, feels a little more catholic than the Pope 17:25
dinner& 17:34
17:52 Altai-man_ joined 17:55 sena_kun left 18:21 AlexDaniel joined 18:22 AlexDaniel left, AlexDaniel joined
AlexDaniel` hmmmmmm 18:26
Geth ¦ old-design-docs: AlexDaniel unassigned from zoffixznet Issue Add warning noting these are now historical documents 18:38
AlexDaniel` Moving some repos, trying to see if something breaks 😇
18:43 domidumont joined 18:57 AlexDaniel left
AlexDaniel` moritz: I don't remember, what happened to ? Do we still need the repo? 18:59
timotimo: what's ? 19:01
jdv79 [Coke]: where? albany?
nine: after i actually pulled your changes they worked. thanks!
moritz AlexDaniel`: no, we can delete that 19:02
timotimo we didn't get far with that, did we ... 19:03
moritz no
timotimo AlexDaniel`:feel free to delete it, i just forked it over to my user 19:04
AlexDaniel` thanks! 19:05
I was just wondering what to do with them
I'm moving some stuff to Raku and some to
also if anybody has a better name for please feel free 19:06
not touching nqp, roast and other important repos for now
first let's see if anything breaks when moving other repos :) 19:07
[Coke] jdv79: albany-ish, yes. 19:10
(about 15m south, across the river)
Xliff nine: Yes. I've made a small change that I think does the right thing. 19:17
nine Cool!
Xliff Still needed directory locking, but that has much less of an impact. Still worrying that that error message means I'm not doing something right, though. 19:18
jdv79 troy? 19:22
Xliff Is S10 hanging for anyone else? 19:28
tony-o AlexDaniel`: you can likely ignore pluggable 19:53
tellable6 hey tony-o, you have a message:
19:53 sena_kun joined
tony-o AlexDaniel`: you available to help me test some stuff using zef as a module maintenance system? (or anyone with a few modules and some time) 19:53
tbrowder: you do need to create manually and then the yaml file - i don't have an auto generator for sqlite yet 19:55
19:55 Altai-man_ left
tony-o .tell tbrowder you do need to create manually and then the yaml file - i don't have an auto generator for sqlite yet 19:55
tellable6 tony-o, I'll pass your message to tbrowder
tbrowder tony-o: thnx! 19:57
19:58 domidumont left 20:03 patrickb joined 20:29 lucasb left 20:51 MasterDuke left
sena_kun totally looks like a bad thing, right? 20:51
ping lizmat ^ 20:52
AlexDaniel` tony-o: I'm not really available right now, and I don't know what you're talking about :T 20:53
lizmat sena_kun: yeah, but caused by me
I guess we need a fallback for .perlseen as well 20:54
sena_kun: give me a min 20:55
AlexDaniel` tony-o: any reason you're not accepting your invite to the Raku org?
tony-o: I noticed when I tried to add you to the reviewers team in the raku org (which is used for giving access to locked tickets and stuff) 20:58
lizmat sena_kun: running spectest with fix now
Geth rakudo: aa9c824d06 | (Elizabeth Mattijsen)++ | src/core.c/Mu.pm6
Some module authors used ".perlseen"

So provide a forwarder for ".perlseen" to ".rakuseen".
lizmat .tell sena_kun aa9c824d06 contains a fix
tellable6 lizmat, I'll pass your message to sena_kun
sena_kun squashable6, status 21:12
squashable6 sena_kun, Next SQUASHathon in 22 days and ≈6 hours (2020-02-02 UTC-12⌁UTC+20). See
Geth rakudo/release-2020.01: 2caee0e964 | Altai-man++ | docs/release_guide.pod
Update new release date
rakudo/release-2020.01: 9d57d35210 | Altai-man++ | 2 files
Log all changes (+ announcement)

Deliberately not logged:
7f15a57 2c5c013 3802cdf 1ec69e1 654d6a1 aa06d32 044b339 88db15f 1d84c64 1d84c64 726c38f d36c2e5 a62d9c7 bbcc342 e03f013 830400f 9528bd0 da03d10 9fa4396 f608a7a 82bd9c2 ... (22 more lines)
rakudo/release-2020.01: 9a215e26b3 | Altai-man++ | 2 files
Remove reverted commit

Deliberately not logged: [c63d8a87]
lizmat sena_kun: you probably want to cherry pick aa9c824d06 for the release
sena_kun lizmat, I am not sure how I'll merge it afterwards, and even if I can cut a release out of some branch that won't be merged, that won't be very cool in terms of build reproducibility... 21:28
anyway, the Blin output looks a bit better, but still has ways for improvement. somehow ADT module was broken. :S
AlexDaniel` sena_kun: what do you mean build reproducibility? 21:46
you can cherry-pick any commit into the release branch
the release branch needs to be merged into master at some point anyway
sena_kun AlexDaniel`, cherry-picking creates a new commit, no?
AlexDaniel` yes, the same commit will exist on master twice
sena_kun I don't want that.
AlexDaniel` it's totally ok
a single commit doesn't hurt, and that's the only way to do it anyway 21:49
once master diverges from the release branch you don't have any other choice, really
like, you have to get the fixes into the release branch and you can't always rebase
sena_kun well, when I can rebase - I rebase. 21:50
AlexDaniel` sena_kun: there was a case when somebody rebased a feature branch incorrectly
sena_kun I suppose it is a matter of preference when there are no "blockers" for either approach.
AlexDaniel` sena_kun: and brought hundreds of duplicated commits when the branch was merged 21:51
ah, sure, if you can, then yeah
at my time the choice was usually 1) cherry-pick and just live with the dup commit 21:52
or 2) bring new untested commits into the release
21:52 Altai-man_ joined
AlexDaniel` first option does no harm besides triggering someone's OCD :) 21:53
Altai-man_ AlexDaniel`, I am kind of ok with bringing more commits, because from my point of view, a relatively complex "fix" can not only fix the regression, but also break some other module, so you need to test everything, and if you have to, why be afraid with grabbing new commits. 21:55
21:55 sena_kun left
AlexDaniel` Altai-man_: yeah 21:55
well, it depends
Altai-man_ anyway, I am starting to create patterns of how to process Blin results, so hopefully I'll master it some day. 21:56
AlexDaniel` not everything can be caught in a timely manner
we just need to fix whateverable to build branches
Altai-man_ would be really awesome to release this weekend, though I am not sure in anything right now
AlexDaniel` and then we'd be able to continuously test the release branch without any issues
Altai-man_ I also find a situation when the module fails in blin, but installs just ok locally, with or without local blin installation. 21:57
tony-o AlexDaniel`: i think i did accept that invite?
AlexDaniel`: oh, apparently i just clicked on it in the email - just accepted it. sorry, thought i'd done that weeks ago 21:58
Altai-man_ this amount of false positives is distracting but I have no idea where to start, because the errors are different.
AlexDaniel` tony-o: yes! Now it's OK!
tony-o: thank you!
Altai-man_: I only cared about bisected results
Geth rakudo: 97e7cdbd54 | (Elizabeth Mattijsen)++ | src/core.c/Rational.pm6
Make Rational.base at least 2.5x as fast

  - completely re-imagined into a multi-method for better optimizing
  - use a native string array to assemble components of the result
  - much faster than 2.5x for values with more fractional digits
  - introducing a :no-trailing-zeroes named parameter.
... (10 more lines)
tony-o pluggable predates the CUR stuff nine did
AlexDaniel` as long as the number of AlwaysFails is low, you don't really need to look at them 21:59
and if it's not low then we're fucked anyway
tony-o AlexDaniel`: do you recall anyone else that was interested in a separate ecosystem stuff? i have a plugin for zef that allows uploading modules to a zef ecosystem and allows zef to search/download/install them from said eco 22:00
AlexDaniel` tony-o: hmmm… not really?
lizmat tony-o: is that a recommendation manager ?
Altai-man_ AlexDaniel`, disregarding Sparrowdo failures (which are reported), I had 320 last time, with false positives like cro.
tony-o lizmat: yes it can double as that 22:01
AlexDaniel` Altai-man_: oh, 320 sounds alright!
Altai-man_: just about 100 more than we should have
Altai-man_ dreams of a better analogue of haskell's slack
lizmat tony-o: can you tell us more about that alternate eco-system ? blog post maybe ?
Altai-man_ AlexDaniel`, is it an irony? :>
tony-o lizmat: for sure, i'll do a write up
AlexDaniel` Altai-man_: what's that? 22:02
Altai-man_: no, no irony, 320 is pretty close
tony-o probably later tonight, i need to ride my bike before the sun gets away
AlexDaniel` Altai-man_: see
Altai-man_: these are pretty much known to fail
Altai-man_ AlexDaniel`, ah, ok, "just 100 more than we should have" sounded a bit playful to me. :)
AlexDaniel` Altai-man_: what's that about hasell's slack? 22:03
Altai-man_ anyway, let's hope a for release this weekend, I plan to dedicate a lot of time then.
AlexDaniel`, sorry, it was a typo of "stack".
AlexDaniel` ah 22:04
Altai-man_ and haskell's stack is cool because they freeze the ecosystem at some point and you know the stuff _WORKS_.
AlexDaniel` nice
(or doesn't, if there are new fixes)
Altai-man_ now when you are making a release and the author of the module breaks something (because take latest from git! is the solution right now), you are like "Hmm, ok, that's not a regression in rakudo, but it is broken now, hmm" 22:05
yeah, if you need something more fresh, you have to wait, that's the downside. :) 22:06
e.g. an author once abandoned a module, another person decided to patch the things and I had to wait like a half of year to get to the new snapshot and to apply some goodies.
AlexDaniel` but then it doesn't show up as a regression in Blin 22:07
Altai-man_ AlexDaniel`, the not cool thing is that Blin can give false negatives. :S
for example, take ADT module
AlexDaniel` yeaaaaaaaaah… 22:08
Altai-man_ it is broken on release branch for sure, it surely works on 2019.11, but what breaks it - no idea.
AlexDaniel` wait but then how is that a false negative?
Altai-man_ AlexDaniel`, ah, sorry, I mean "false negative" in "doesn't show up as a regression, just as a failure sense", not the "it passes, but is in fact broken" sense. 22:09
kind of harder when you have more than two states.
ok, I am running away to get some sleep
AlexDaniel` hmm
Altai-man_ afk& 22:10
Geth roast: b3cd406ea2 | (Elizabeth Mattijsen)++ | S32-num/base.t
Fix erroneous test

Since this test would fail. Since the test involved *no* specific number of digits, any trailing 0 should be removed. The new implementation does that in this case, the old implementation of Rational.base did not.
lizmat sleep& 22:59
23:53 sena_kun joined 23:55 Altai-man_ left