|
00:16
abraxxa-home left
|
|||
| guifa | aha | 00:19 | |
| I can do uv_udp_getsockname(), then read the sin6_scope_id | 00:20 | ||
| Just annoying because i have to convert to an sockaddr, extract the interger, then reformulate into a str | 00:25 | ||
|
01:03
MyNetAz left
|
|||
| antononcube | .seen lizmat | 01:07 | |
|
01:18
MyNetAz joined
|
|||
| guifa | oh fun | 02:03 | |
| Undefined symbols for architecture arm64: "_uv__idna_toascii", referenced from: _uv_getaddrinfo in libuv.a[18](getaddrinfo.o) | |||
|
02:04
hulk joined
02:05
kylese left
|
|||
| ugexe | We don’t build libuv normally and instead include the files ourselves in a makefile or some such. Often the files (or order of those files) needs to be changed to use features/code paths that weren’t used yet | 02:06 | |
| guifa | that's what I was gathering -- was trying to see where I'd be able to link it in | 02:09 | |
| is it as simple as adding in `3rdparty/libuv/src/idna@obj@` where I see all of the other similarly named lines? | 02:10 | ||
| answer: it seems so, compile was successful, time to try it out | 02:14 | ||
| whelp, it works, but apparently uv_udp_getsockname doesn't write back to the scope_id | 02:50 | ||
| there has to be an easier way to figure out the ipv6 interface from a bound socket | |||
|
03:01
hulk left,
kylese joined
|
|||
| ugexe | have you tried looking at node or julia source code? | 03:02 | |
|
03:03
Aedil joined
|
|||
| ugexe | llm might even be able to somewhat correctly guide you on libuv usage | 03:04 | |
| guifa | Yeah I think I'm going to have to do a deep dive on Node, that's so far been the solution to my issues | 03:10 | |
|
03:15
kylese left,
kylese joined
|
|||
| guifa | From what I can see with Node.js, it looks it's unavoidable to have to effectively manually specify the interface | 03:15 | |
| At least, not without doing a lot of potentially platform specific hoops to determine the prefered/default interface (seems like it's supposed to be index 0, but on my system, the "default" is at idx 12, with no interfaces at 0, and loopback at 1) | 03:20 | ||
| I guess the easiest for now is on the Raku side to make it be IO::Socket::Async.bind-udp($addr, :multicast($interface)) | 03:22 | ||
| and leave it to module space to figure out the ideal one | |||
| Julia doesn't seem to require an interface. Trying to track down their code | 04:39 | ||
| guifa thinks they might be listening on all interfaces, which is certainly one solution | |||
| Nope. Both just pass a NULL to lib_uv for the interface. That just clearly doesn't work on macOS :-) | 06:44 | ||
|
06:51
gfldex left
07:36
Sgeo left
07:40
MyNetAz left
07:55
MyNetAz joined
08:07
acidsys left
08:30
acidsys joined
09:51
melezhik joined
|
|||
| melezhik | Dormitory post has got some UPs (+1 for just few hours) on programming.dev - programming.dev/post/23334003 - please support if you wish ) | 09:52 | |
| +11 | 09:53 | ||
|
09:56
melezhik left
09:57
MyNetAz left
10:12
MyNetAz joined
10:15
sena_kun joined
13:15
Aedil left
13:20
Aedil joined
|
|||
| antononcube | @guifa What you are trying to find out looking at Julia? How to implement a DNS service? And / or socket library? | 13:24 | |
| Sorry, I am too lazy to read all of the related discussion above… | |||
| guifa | antononcube: looking at how they implement IPv6 multicast, because their sample code didn't seem to be too terribly bothered with specifying network interfaces, which the underlying libuv (which Raku also uses) effectively requires | 13:25 | |
| antononcube | I wonder then how Mathematica and R do that. (Same type of languages.) | 13:26 | |
| guifa | github.com/Raku/problem-solving/issues/457 <-- bottom post sums up the current state of affairs | ||
| I can create something that works regardless how we want to, but the question is how to wire it up API wise really | 13:27 | ||
| antononcube | But, more importantly for me, I can / should hook up a Julia kernel to Jupyter. Also, at some point I experimented using Julia in Raku sessions. | 13:28 | |
| “bikeshedding” 🙂 | 13:30 | ||
| guifa | lol maybe Is hould just change it to "Multicast API" or something | 13:33 | |
| timo | oh no, the bike is shedding! | 13:35 | |
|
13:36
tadzik left
|
|||
| antononcube | Some of that discussion reminds of designs in ZeroMQ. | 13:59 | |
| But using ZMQ for inspiration might be too much of a stretch. | |||
| timo | for the purpose of what guifa is doing, there is a hard limit on the amount of innovation we can do, since it's meant to interoperate with other regular programs that use this kind of multicast | 14:01 | |
|
14:01
tadzik joined
|
|||
| timo | we can make the API more convenient to use for as many use cases as possible, but we're tied to the semantics of existing UDP multicast system calls and such | 14:03 | |
| guifa | ZMQ is definitely a bit higher level than what I'm trying to work out | 14:16 | |
|
14:16
sena_kun left
|
|||
| timo | and cro has support for ZMQ right? i haven't tried it yet, though. | 14:16 | |
|
14:18
sena_kun joined
|
|||
| guifa | I'm happy at least I've figured out the lower level (MoarVM) issues, though. Once I know/decide how the higher-level side should look, I can clean up/adjust MoarVM to follow. | 14:18 | |
| "figured out" eq "finally groks everything going on under the hood" | 14:19 | ||
| timo | .o( boss, we caught another one! ) | ||
| guifa | lol not saying I get MoarVM's internals fully, just enough to handle the socket setup and libuv :) | 14:22 | |
| timo | not long ago i realized i don't thoroughly grok the async event loop thread and "active work" and such | ||
| there's still a bug related to doing stuff in a bad sequence that one of our spec tests flaps with that i couldn't really debug to the end | 14:24 | ||
| guifa | although seriously, the write up on this is going to be legendary. "So multicast didn't seem to work. Let's try to figure out why. Here's the Rakudo code. And now the NQP, and now the MoarVM. But now to understand the MoarVM, we need to understand libuv. Oh, and study back on up on oddities with IPv6. Let's now compare with NodeJS. Oh, welcome to cross-platform issues. Okay, phew, all that's done. Oh wait, how will we enable all of this from Raku? | 14:25 | |
| 🤦♂️" | |||
|
14:34
Sgeo joined
|
|||
| antononcube | @guifa Maybe you should use some of the reasoning LLMs. | 18:51 | |
| I would say, there is 70% probability they would help. | |||
|
20:52
Aedil left
22:37
sena_kun left
|
|||
| lizmat | news.ycombinator.com/item?id=42518624 | 23:17 | |
| [Coke] | how are they getting a startup time of 15s? | 23:21 | |
| coleman | This is anecdotal, but I used the XML library for the first time yesterday and the first time I ran my script it was slow. Subsequent runs were fast. | 23:31 | |
| Is there some compilation step that a totally fresh install would encounter? | |||
| timo | yeah, precompilation of the module | 23:59 | |