github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
00:19 timotimo left 00:20 timotimo joined 00:44 Kaeipi joined 00:47 Kaiepi left 01:11 squashable6 left 01:12 squashable6 joined 02:02 MasterDuke left 05:46 Geth left, timotimo left, timotimo joined 07:50 patrickb joined 08:00 domidumont joined
nwc10 good *, #moarvm 08:11
3 months later, still no update: blog.pyston.org/
(monetising infrastructure assumed to be free is hard. Let's go shopping) 08:12
moritz good morning, nwc10, * 08:35
08:35 Geth joined
moritz yes, maintaining a separate compiler/JIT/VM is really quite some work, thankless and hard to get paid for 08:36
there are several such projects in the python world, and whenever I try them, they don't quite work for my use case 08:37
nuitka is one of those
nwc10 I've heard of that one, but (like all of them) not tried it, and I forget exactly what it does. Static compiler to C? 08:38
IIRC it's the one that Guido has dumped on. At least, the only one that I remember that Guido actually noticed enough to say that he thought something about it was not actually a sustainable idea. (But I forget why or when) 08:39
moritz yes, that's the python to C compiler 08:44
and it just had another release 4 days ago. Noice 08:46
09:02 zakharyas joined
jnthn good * 10:19
So...can I assemble the new coffee machine before it has produced a cup of coffee?
nwc10 this is a bootstrapping problem? 10:28
jnthn Pretty much 10:29
nwc10 in normal times this would not a problem - you'd just pop out to a nearby cafe for a "DR site" coffee 10:31
jnthn OK, there is coffee in my cup. It's only half full, and I can't yet figure out how to resolve that. 10:57
nwc10 redundant array of inexpensive cups?
just iterate on half cups until you find a solution? 10:58
find inspiration, I meant
jnthn The original one was a lever-style one and I just had a water on/off switch and waited until the cup was full 10:59
10:59 squashable6 left
jnthn This is one of the fancy automatic ones that I can - once I finish my supply of ground coffee - put beans in and get freshly ground. But it wants to own the whole process. :) 11:00
Hm, this does paste pretty good at least.
nwc10 no, don't *paste* coffee. It will ruin your keyboard
jnthn oops, *taste 11:01
apparently I need to drink more of it :)
nwc10 yes.
have you firewalled it yet?
11:02 squashable6 joined
jnthn I...I don't think it tries to connect to the internets. But given half the washing machines I found when I was shopping for one of those last year did, maybe I should be worried... :) 11:04
Why do I want my washing machine online?!
nwc10 so that someone else can mine bitcoin with it!
well, d'oh!
jnthn OK, so finally I'm maybe ready to write some code... :) 11:06
nwc10 I was misremembering. CVE-2017-7240 was a *dishwasher*, not a washing machine.
11:07 Kaeipi left
nwc10 oh, actually it's more fun than that. I hadn't read this official grovelling admission before: www.miele.de/en/m/miele-admits-com...h-4072.htm 11:08
jnthn aha, I found how you configure cup size, so I can get a full cup soon :)
11:09 Kaeipi joined 11:20 Kaeipi left 11:21 Kaeipi joined 12:01 zakharyas left 12:09 Geth left 13:00 cog__ left 13:01 cog__ joined 13:37 lucasb joined
[Coke] whew 13:40
13:45 ggoebel joined
ggoebel our coffee machine recently gave up the ghost too. Only 5 years old. Our experience has been that the more complicated the coffee machine is, the shorter it lasts. We now use a hot water kettle and french press. 13:48
jnthn Darn, this new one is more complicated than the 13-year one it replaces :) 13:51
[Coke] I do miss the "going to work" part of the day where you could pick up a coffee on the way (I switched to mostly work from home even before the pandemic)
jnthn I suspect the old one may well be fixable by somebody with some sort of clue what they're doing (or me if I researched it a bit); it seems like a rubber seal has decayed 13:55
OK, so..."just" the stateful bit of dispatch resumptions to go, and then in theory I have enough to write a test case for a dispatcher doing basic method deferral 14:53
jnthn wonders how long until he sees the light at the end of the new-disp tunnel... 14:54
15:16 zakharyas joined 16:00 linkable6 left 16:01 linkable6 joined 16:33 samcv left, samcv joined 16:53 linkable6 left, linkable6 joined 17:25 domidumont left
nwc10 nine: I think I have it figured out. If you look at the camera lean on www.youtube.com/watch?v=hpqoqJxbNpw&t=0s currently - when SN9 "fell over", it was only trying to get vertical. 18:58
19:47 MasterDuke joined
MasterDuke www.sba-research.org/wp-content/up...coop10.pdf i just gave it a quick skim, but it looks kind of like the specializations spesh does? 19:48
20:16 MasterDuke left 20:18 MasterDuke joined 20:22 Kaeipi left 20:23 Kaeipi joined
jnthn MasterDuke: I've looked through that one before, and yes, it's in the same solution space 20:35
MasterDuke ah ha, glad i'm not completely off base 20:46
20:58 Kaeipi left 20:59 Kaeipi joined 21:00 zakharyas left
jnthn is gradually starting to feel comfortable with the dispatch resumption design that he's implementing 21:29
The pieces I've put in place today have each worked first time (silly thinkos aside, but really I mean there weren't any unexpected design issues). 21:30
And I can see how they'll handle the gnarly stuff like nextcallee and lastcall
nwc10 [Coke]: actually, my definition wasn't quite accurate - Wet Dress Rehersal is where you test everything about your launch process, right up to the point of (not) lighting the engines. With real fuel. 21:33
side effect (obviously) is that you also get to test the "how to make it safe again" process that you'd have if your (for real) launch aborts before ignition. 21:34
jnthn I'm also fairly happy that in a steady state of deferrals it can be allocation free, and that it will let us do the calculation of the method to defer through once-ish per `callsame` site and cache it, rather than having to calculate it every time. 21:35
Plus that the dispatch programs it stacks up at the inline caches will provide information that will let us do specialization linking and, where size allows, inlining.
A couple of unknowns on inlining still remain, though nothing that overly worries me 21:36
nine nwc10: that's an.....explanation :D
MasterDuke jnthn: so mostly a smop left? 21:48
22:03 El_Che joined
El_Che Hi, I am able to build moarvm on 20+ distro/version combinations, except on Fedora Rawhide (the dev, next-to-be-latest OS): github.com/nxadm/rakudo-pkg/runs/1...step:5:308 22:04
Now building rakudo on the released fedoras 31 and 32 22:05
MasterDuke El_Che: it looks like the 3rdparty/libatomicops submodule didn't get clones/pulled somehow? 22:10
El_Che no 22:11
I download the tag.gz in an action before running the container
and all the containers get the same tar.gz
(a copy)
MasterDuke huh. then i really couldn't say 22:12
El_Che same here
:)
fedora 31 builds fine 22:14
22:22 raku-bridge2 joined 22:23 raku-bridge2 is now known as raku-bridge 22:33 linkable6 left 22:37 squashable6 left
jnthn No guesses either, fwiw. It's the same tarball used over all distros? Is it reproducible? 22:46
22:46 bisectable6 left, coverable6 left 22:47 statisfiable6 left, quotable6 left, committable6 left, unicodable6 left, greppable6 left, notable6 left, evalable6 left, sourceable6 left, shareable6 left, benchable6 left, tellable6 left, nativecallable6 left
jnthn MasterDuke: There's still some non-simple things to go, but at least by now I feel I'm making progress; when I last looked at this in the autumn I went through a bunch of designs that each didn't work 22:48
Even on paper, let alone in implementation
22:49 evalable6 joined, benchable6 joined, squashable6 joined 22:50 shareable6 joined, sourceable6 joined, nativecallable6 joined, statisfiable6 joined, bisectable6 joined 22:51 bloatable6 joined 22:52 releasable6 joined
El_Che jnthn: same file, reproducible 22:53
my github flow is:
- step 1: download the sources and keep them as artefacts 22:54
- step 2: in parallel build the source for all the distros
22:55 statisfiable6 left, squashable6 left
El_Che (fedora 32 builds fine as well) 22:55
22:55 benchable6 left, nativecallable6 left, sourceable6 left, shareable6 left, evalable6 left 22:56 releasable6 left
El_Che so something changed between distros 22:56
(we've seen it before in the past) 22:57
22:57 bloatable6 left, bisectable6 left 22:58 notable6 joined
El_Che I am trying to add the next-stable releases so we can catch things like that early on 23:00
23:00 bisectable6 joined, nativecallable6 joined, statisfiable6 joined, unicodable6 joined, benchable6 joined, squashable6 joined, committable6 joined, shareable6 joined
jnthn El_Che: Thanks, it's really good to do that. 23:01
El_Che: I'm not surprised that things change between distros. I *am* surprised about this particular failure mode 23:02
El_Che I told lizmat earlier, that the idea to allow to run it with commits for every components as well, eg moarvm/nqp releae 2020.12 and rakudo@acdsomehash 23:03
fork the repo, edit the version file and push and the action will run 23:04
work in progress :)
23:04 bisectable6 left, nativecallable6 left, shareable6 left, benchable6 left, unicodable6 left, squashable6 left, statisfiable6 left, committable6 left, notable6 left 23:11 tellable6 joined, benchable6 joined, nativecallable6 joined, bisectable6 joined, notable6 joined 23:12 committable6 joined, bloatable6 joined, releasable6 joined, quotable6 joined, squashable6 joined, statisfiable6 joined
El_Che jnthn: alpine edge (dev release) fails earlier while the released alpines build fine: github.com/nxadm/rakudo-pkg/runs/1...focus=true 23:13
I think something like new coreutils
I have seen it before
when errors don't make sense, look at the build tooling :) 23:14
jnthn hah, that's an interesting one :d 23:15
*:D
El_Che I meant binutils
I did this for the 2020 release:
23:15 benchable6 left, bisectable6 left, tellable6 left, nativecallable6 left
El_Che # binutils bug 23:15
wget \
de.archive.ubuntu.com/ubuntu/pool/m..._amd64.deb && \
dpkg -i *.deb && rm -rf *.deb &&
23:15 notable6 left
El_Che retrieved deb from an older ubuntu release 23:15
23:17 statisfiable6 left, committable6 left, bloatable6 left, releasable6 left, quotable6 left, squashable6 left 23:19 bisectable6 joined, benchable6 joined, quotable6 joined, committable6 joined, notable6 joined, statisfiable6 joined, squashable6 joined
El_Che all builds with the 2 aforemetioned disabled: github.com/nxadm/rakudo-pkg/action.../522238929 23:21
23:22 bloatable6 joined
El_Che all green, I like that :) 23:25
23:30 benchable6 left, bisectable6 left, bloatable6 left, notable6 left 23:31 quotable6 left, committable6 left, statisfiable6 left, squashable6 left 23:46 ggoebel left 23:48 ggoebel joined 23:49 benchable6 joined, bisectable6 joined, committable6 joined, notable6 joined, squashable6 joined, statisfiable6 joined, bloatable6 joined, greppable6 joined, nativecallable6 joined, quotable6 joined, coverable6 joined, tellable6 joined 23:51 linkable6 joined, shareable6 joined, releasable6 joined, sourceable6 joined, unicodable6 joined 23:59 evalable6 joined