00:01 sena_kun joined 00:03 Altai-man_ left 00:55 Xliff joined 01:30 softmoth joined 02:01 Altai-man_ joined 02:03 sena_kun left 02:21 softmoth left 02:22 softmoth joined 04:02 sena_kun joined 04:03 Altai-man_ left 04:53 sena_kun left 04:54 sena_kun joined 06:01 Altai-man_ joined 06:04 sena_kun left 06:49 Xliff left 06:57 Kaeipi left 07:00 Kaiepi joined 07:11 softmoth left
nine AlexDaniel: hello? 07:45
MasterDuke nine: i think he was pinging about github.com/rakudo/rakudo/issues/3769 07:53
08:01 sena_kun joined 08:03 Altai-man_ left 08:13 JJMerelo joined
timotimo how come nothing ever exploded due to missing frames? 08:26
they are only present, but not useful, perhaps?
nine yes 08:45
I.e. they are only run during compilation
timotimo ah, bummer that we can't get rid of them 08:55
unless we can, somehow 09:00
09:12 JJMerelo left
nine We probably could. But why invest the time? 09:18
timotimo i cry a little inside when i think of all the times i tried to make core setting size smaller :D 09:20
nine Making the stuff we read smaller certainly helps as we have to read less. But stuff that we don't actually read shouldn't cause any issues and indeed I haven't seen any benchmarks getting worse 09:21
timotimo right. just a thing of install size 09:23
nine Which hasn't been a topic previously 09:52
10:01 Altai-man_ joined 10:03 sena_kun left 10:42 JJMerelo joined
AlexDaniel whatever 11:28
you want to slap 13M on top of the installation, feel free
I'm just maintaining the bots
11:44 patrickb joined
Altai-man_ I suspect this change wasn't done on a whim "Oh let's add something into our precomp files", but has technical reasoning behind it. 11:52
11:56 lichtkind joined
nine The gist is: I had good reason for the change. I'm pretty sure we could reduce the size gains or even eliminate them completely. I'm also sure it's going to be complicated and cause a couple of headaches. 11:58
Thus I think my time is better spent on in-process precompilation as that seems to be able to cut compilation times by 1/3 for multi-file code bases.
lizmat nine++ 12:01
12:01 sena_kun joined
ShimmerFairy space is pretty cheap nowadays, I'd be shocked if rakudo's size became a serious issue. 12:02
timotimo we do have one build for every commit 12:03
12:03 Altai-man_ left
nine It can matter on embedded systems like WiFi routers. But on mine, where I run FHEM for home automation which is written in Perl, even perl is too large and I had to move it onto an attached disk. 12:03
ShimmerFairy That's true, I forgot about embedded systems. .oO(and I suppose space isn't all that cheap on SSDs either) 12:06
lizmat timotimo: bzipping reduces by 50+% ? 12:07
timotimo we use zstd; would have to ask AD how the builds compare compressed 12:08
nine Nah, even if you use super fast storage, you end up at about 120 Euros per TB. So storing a 57 MB rakudo installation on super fast NVME costs about 0.7 Euro Cents. 12:09
ShimmerFairy Last I checked, HDDs are still incredibly cheap for the storage. 12:13
nine Yeah, they cost practically nothing 12:14
12:28 JJMerelo left 12:45 Kaiepi left 12:48 Kaiepi joined
Kaiepi gonna start testing out my dns changes on more platforms once windows update finishes up 12:51
they perform fairly well on my own machine, but not so much my freebsd box
could be a configuration thing or st, not quite sure until i see how it does on linux and windows 12:52
13:02 Kaeipi joined 13:06 Kaiepi left 13:13 Kaeipi left
AlexDaniel timotimo: well, you can check it yourself 13:49
shareable6: HEAD
shareable6 AlexDaniel, whateverable.6lang.org/HEAD
AlexDaniel shareable6: 537f8877034^
shareable6 AlexDaniel, whateverable.6lang.org/537f8877034^
AlexDaniel that's a difference of 0.5M per build 13:50
but then, for older stuff, multiple builds are compressed with long-range compression, so it's possible that it doesn't matter at all 13:51
for me that is 13:52
but then if you want to talk about it, then first get me a server with this cheap magical 1 TB SSD, and then we can have a discussion 13:53
I know I can buy it for my own computers, but I'm not using my own computers to run whateverable 13:54
sena_kun: it's not about doing it on a whim, but we have to acknowledge that nobody knew that this change increases installation size by 13M 13:55
13:57 softmoth joined
sena_kun AlexDaniel, fair enough. I suspect dispatcher overhaul can have some positive influence on space, since bytecode promises to be shorter. No idea how much until it's done, though. 13:58
13:58 lucasb joined
AlexDaniel thing is, you can make it 1 GB per installation and then claim that the space is cheap, I don't care anymore 13:59
nine sena_kun: my guess, a few %, nothing major
AlexDaniel: what machine does whateverable run on?
AlexDaniel nine: it's my server 14:00
well, a dedicated server I pay for
14:01 Altai-man_ joined
Altai-man_ nine, well, then not so much. I was under an impression raku bytecode is heavy on calls, though maybe it is not _so_ heavy considering inlining and other tricks. 14:01
nine AlexDaniel: my server's got some 750G unclaimed space left. Would that help with anything?
AlexDaniel nine: Xeon E3-1245 V2, 3x 120GB SSDs in RAID1, 16GB of RAM
bots take around 10GB of RAM I think 14:02
and less than 40GB of storage
if that's 750G of space on SSDs then it will make bisectable a tiny bit faster
14:03 sena_kun left
AlexDaniel (because I won't be compressing builds as aggressively) 14:04
if you're willing to run the bots then I can perhaps dockerize them a bit better so that it's easier for you 14:08
nine Sadly it's still HDD. It's an i7-2600 CPU @ 3.40GHz, 32 GB RAM and 2x3T hard drives. Really can't complain for 32,27 Euros per month. 14:10
AlexDaniel mine is also 32€ per month 14:12
14:21 b2gills left 14:55 Xliff joined, Kaiepi joined 14:56 Xliff left
Geth rakudo/match-refactor: 11 commits pushed by (Elizabeth Mattijsen)++
review: github.com/rakudo/rakudo/compare/1...677f337de6
15:02
lizmat hates branches 15:03
Altai-man_ lizmat, o/ 15:11
lizmat Altai-man_ o/
and? Blin results ?
Altai-man_ gist.github.com/Altai-man/757af6b6...aa35641767 <- the match-refactor branch output. There are quite a lot of modules, so pick one which sounds the easier.
lizmat Altai-man_: thanks! 15:12
Altai-man_ Yup, just recalled that because of the branch notion, guess they are not so bad. :)
You are welcome, ping me anytime if anything else can be be helpful.
lizmat I will!
15:16 b2gills joined 15:27 Kaiepi left
AlexDaniel Altai-man_: oh, that's Blin running for a branch? It works? 15:27
awesome
I wasn't entirely sure if I actually made it work
let me know if there are any issues 15:28
Altai-man_ AlexDaniel, yeah, just worked fine thanks to your patches. :)
AlexDaniel, sure! Thank you for working on it.
AlexDaniel
Geth rakudo/match-refactor: f49982c991 | (Elizabeth Mattijsen)++ | src/core.c/Match.pm6
Apparently, we can directly work with Cool values

This fixes Template::Mustache
15:29
AlexDaniel c: match-refactor say 42 15:31
committable6 AlexDaniel, ¦match-refactor: «42␤»
AlexDaniel lizmat: you can also do that ↑
c: match-refactor say $*PERL.compiler.version 15:33
committable6 AlexDaniel, ¦match-refactor: «v2020.06.14.g.19.f.0.f.53.e.0␤»
AlexDaniel hmm I don't know what it does for force pushes 15:34
but please start using it and tell me about the problems, I'll fix it
lizmat AlexDaniel: thanks, but for me it's more important to run things on HEAD to compare, so I just use camelia for that privately 15:37
AlexDaniel hmmmmmmmmmmmm 15:38
lizmat: what if evalable was answering your private messages?
and committable
lizmat that would be nice 15:39
AlexDaniel lizmat: do you have a cloak?
lizmat: e.g. try `/whois AlexDaniel` it'll say (~user@perl6/developer/alexdaniel) 15:40
lizmat don't think I have one ? 15:41
AlexDaniel mst: can you give a raku/developer cloak to lizmat?
mst: moreover can you change existing ones from perl6 to raku?
15:50 Kaiepi joined 16:01 sena_kun joined 16:03 Altai-man_ left 16:24 JJMerelo joined
Geth rakudo/match-refactor: 3a37f876c7 | (Elizabeth Mattijsen)++ | src/core.c/Match.pm6
Make the .from and .to method private

This is one of the disadvantages of having merged the Cursor into the Match. Because .from is now a HLL method, it was shadowing a rule with the same name in any grammar.
Match eqv Match now checks from/to attribute directly.
Fixes Docker::File.
16:29
16:34 Kaiepi left
AlexDaniel c: match-refactor say 42 16:58
committable6 AlexDaniel, ¦match-refactor: «42␤»
AlexDaniel c: match-refactor say $*PERL.compiler.version 16:59
committable6 AlexDaniel, ¦match-refactor: «v2020.06.27.g.3.a.37.f.876.c␤»
AlexDaniel oh, it's working
nice
btw, one interesting thing the bots can do
is
bisectable6: vars
bisectable6 AlexDaniel, Will bisect the whole range automagically because no endpoints were provided, hang tight
AlexDaniel :|
okay maybe it doesn't work after all
committable6: vars
committable6 AlexDaniel, timeout=10
bisectable6 AlexDaniel, ¦6c (44 commits): «04===SORRY!04=== Error while compiling /tmp/KwIPDlQ3cP␤Undeclared routine:␤ vars used at line 1. Did you mean 'VAR'?␤␤ «exit code = 1»»
AlexDaniel, Nothing to bisect!
AlexDaniel committable6: timeout=50 17:00
committable6 AlexDaniel, timeout is now set to “50” (default value is “10”)
AlexDaniel it used to be half-broken because of the %*BOT-ENV thing, but with an upcoming commit it'll be fixed
bisectable6: timeout=20
bisectable6 AlexDaniel, Will bisect the whole range automagically because no endpoints were provided, hang tight
AlexDaniel okay some kind of priority issue 17:01
bisectable6 AlexDaniel, ¦6c (44 commits): «04===SORRY!04=== Error while compiling /tmp/JrC13tlJA3␤Undeclared routine:␤ timeout used at line 1␤␤ «exit code = 1»»
AlexDaniel, Nothing to bisect!
AlexDaniel releasable6: vars
releasable6 AlexDaniel, branch=master; timeout=10
AlexDaniel sena_kun: ↑ for example, this way you can set releasable to complain about some other branch (release branch) instead of master 17:02
let's actually see if this works 17:03
evalable6: vars
evalable6 AlexDaniel, commit=HEAD; timeout=10
AlexDaniel evalable6: commit=match-refactor
evalable6 AlexDaniel, commit is now set to “match-refactor” (default value is “HEAD”)
AlexDaniel evalable6: say $*PERL.compiler.version
evalable6 v2020.06.27.g.3.a.37.f.876.c
sena_kun Hm-hm, this is very interesting.
AlexDaniel lizmat: ↑
so evalable6 is one command away from tracking a different branch
lizmat :-) 17:05
AlexDaniel and any constant can be easily turned into a variable, so we can make these bots as configurable as you want
evalable6: commit=HEAD 17:06
evalable6 AlexDaniel, commit is now set to “HEAD” (default value is “HEAD”)
AlexDaniel actually
evalable6: reset commit 17:07
evalable6 (exit code 1) 04===SORRY!04=== Error while compiling /tmp/tMWMbf45Wo
Undeclared routines:
commit used at line 1
reset used at line 1
AlexDaniel hmm don't remember how to do that :D
evalable6: commit=reset
evalable6 AlexDaniel, commit is now set to its default value “HEAD”
AlexDaniel ah, this is how
the changes only last till the bot is restarted, so don't worry about breaking anything 17:09
17:19 MasterDuke left 17:59 JJMerelo left 18:01 Altai-man_ joined 18:03 sena_kun left 20:01 sena_kun joined 20:03 Altai-man_ left 20:35 finsternis joined
Geth rakudo/match-refactor: 47124c44e9 | (Elizabeth Mattijsen)++ | src/core.c/INTERPOLATE.pm6
Simplify !cursor_pass logic in INTERPOLATE

  - only do the bare essentials, which is all what INTERPOLATE needs
21:10
rakudo/match-refactor: ab75e253d8 | (Elizabeth Mattijsen)++ | src/core.c/INTERPOLATE.pm6
Move lack of support for hashes into a separate candidate
21:22
21:30 softmoth left 21:33 softmoth joined 21:40 softmoth left
Geth rakudo/match-refactor: ee0174ac9f | (Elizabeth Mattijsen)++ | src/core.c/INTERPOLATE.pm6
Turn a condition around

Saves 1 not per INTERPOLATE
21:45
22:01 Altai-man_ joined 22:03 sena_kun left
Altai-man_ lizmat, want another blin run? 22:03
lizmat yes, please :-) 22:04
it will still contain fails, like Inline::Perl5, but others may be fixed other than the ones I specifically checked 22:05
Altai-man_ started
lizmat Altai-man++
Altai-man_ yes, think will help to narrow things down a bit
22:12 nebuchadnezzar left 22:48 lucasb left 23:35 nebuchadnezzar joined 23:47 patrickb left