00:37 eseyman left 00:47 rmv left 01:29 guifa left 01:48 hulk joined 01:49 kylese left 01:50 guifa joined 02:09 guifa left 02:15 hulk left, kylese joined
simon_sibl glad I can help with my little knowledge to find issues xD 02:15
02:42 rmv joined 02:47 rmv left
timo what's "the memory issue"? 02:47
03:59 lichtkind__ joined 04:02 lichtkind_ left
simon_sibl this ? github.com/rakudo/rakudo/issues/6368 04:11
timo ah. i would have expected that to go away with locking. i'll look into it to make sure it's not a moarvm bug but none of the changes in the last two months looked like they could cause something like that 04:14
does it always happen? after some amount of time or pretty much immediately? 04:16
any difference between a fresh from-zero download or a resume? 04:17
> Done Downloading !!! 04:21
simon_sibl it seems to be random 04:23
but with the latest version, it either crashes with that message or just get stuck immediately
timo yeah that getting stuck should be the utf8 decoding thing
simon_sibl but I thought 2026.06 fixed it ? 04:24
timo no, 2026.06 introduced it, a point release for 2026.06 could fix it 04:27
simon_sibl ah I see
so which version did you use now to test the download ?
timo is it important that bstr in the Bencode actions sometimes returns a Str? 04:30
that's where it hangs for me
you can try utf8-c8 as the encoding there, it will never fail to give you a Str, though invalid utf8 will look funky
m: my $in = Buf.new(0x22, 0x98, 0x06, 0xbc, 0xb8, 0x3a, 0x96, 0xd9); say $in.decode("utf8-c8"); say try $in.decode; say $! 04:33
camelia "􏿽x98􏿽xBC􏿽xB8:􏿽x96􏿽xD9
Nil
Malformed UTF-8 near byte 98 at line 1 col 2
in block <unit> at <tmp> line 1
timo i guess making it decode as utf8-c8 immediately gives me "done downloading!" at 0/0 pieces, hehe. 04:35
i can't get it to give the moarvm oops over here 04:39
i hope seeders don't get mad at me for restarting the download so often 04:41
what operating system are you on? if you're on linux, you can install `rr` and make me an `rr record`ing so i don't have to try to reproduce it locally myself 04:52
if you're using rakudobrew to get your rakudos, you can ask it to build a moarvm from commit a3a507141 04:56
linkable6 (2026-06-30) github.com/MoarVM/MoarVM/commit/a3a5071413 utf8: Do not lock up when very long string has invalid utf8 in it.
timo shareable6: 1e6543d83 04:57
shareable6 timo, whateverable.6lang.org/1e6543d83
timo you can also use this pre-build rakudo which will have the right moarvm in it
you seem to be protecting the pieces manager with different locks, that doesn't seem like a good idea? 05:06
oh you pass the pm-lock to BT::Peers.new, never mind 05:07
05:08 manu_ joined, manu_ is now known as eseyman
timo even with the protect calls to the lock I can't get the crash 05:12
simon_sibl let me try to use rakubrew and the commit and then rr 06:09
hmm 06:16
timo i'm thinking about what can be done to give helpful information when a "pointer to past fromspace" is encountered. it has to be something that doesn't impact performance otherwise; i can think of ways to make it spit out helpful information if it's fine to use a bunch more memory and be a bunch slower :P 06:18
simon_sibl is it this way to build with the right moarvm commit: rakubrew triple main main a3a5071413 06:19
linkable6 (2026-06-30) github.com/MoarVM/MoarVM/commit/a3a5071413 utf8: Do not lock up when very long string has invalid utf8 in it.
simon_sibl otherwise not I am building moar-blead 06:20
06:21 rmv joined, rmv left, rmv joined
timo if you use "main" for rakudo, it will already pull in the moarvm with the patch for free 06:22
since we have bumped the dependencies for nqp and moarvm after that commit landed
simon_sibl so triple main main main 06:23
I will see how that goes, I built moar-blead but not sure how to actually use it with rakubrew 06:24
06:25 rmv left
timo i assume just "rakubrew switch moar-blead"? 06:26
i don't use rakubrew myself
AFKBBL 06:27
simon_sibl well with moar-blead, I dont get the freeze it seems 06:38
Voldenet: your code for the supply analyzer looks very nice ! 06:41
just get issues with the Buf stringification 06:43
06:47 Sgeo_ joined
I modified the $value to be printed as $value.raku, it spits a lot of values but it works xD 06:48
but yeah so far with the latest version, no freeze 06:49
06:51 Sgeo left 06:57 rmv joined, rmv left, rmv joined 07:04 rmv left
timo I can recommend ".gist" instead of ".raku" which in many cases will shorten the output some 07:07
07:43 abraxxa joined, Sgeo_ left 07:50 abraxxa left 07:51 abraxxa joined
Voldenet Yeah, this analyzer is basically giving you events, to make customizing easy – you could also capture some context for supply inspector (like some id), count over time aggregates, write it to sqlite etc. 08:02
initially I put the prints inside INSPECT-SUPPLY but quickly realized that the output is horrible and that there was no way to easily improve it 08:03
collected stats could be also pushed into prometheus and visualised in grafana, so it'd be possible to say "what eats so much cpu" without analyzing large strings log 08:06
timo I've been meaning to look more closely at the format that the firefox profiler tool eats 08:12
we have Log::Timeline, but the only frontendfor that currently is in Comma 08:13
Voldenet Hm, that is potentially useful and could be integrated with supply inspection 08:25
08:45 rmv joined, rmv left, rmv joined 08:49 rmv left
lizmat weekly: raku.foundation/Donations 08:57
notable6 lizmat, Noted! (weekly)
timo yeah it can eat point-in-time samples with stacks, and it can eat points with details in "markers" as well as start-end things in the markers IIUC 09:01
github.com/mstange/samply/issues/349 it seems like you can write a very simple "markers-$PID.txt" file that samply will pick up, it can do ranges as well as point-in-time events 09:47
10:44 abraxxa left 12:01 lichtkind__ left 12:13 lizmat left, lizmat joined
timo `marker-$PID.txt` actually, as it turns out (and on linux it will only be picked up if there's a corresponding mmap event for that file) 12:14
12:49 rmv joined, rmv left, rmv joined 12:54 rmv left 13:03 apogee_ntv joined 13:27 librasteve_ joined 13:31 rmv joined, rmv left, rmv joined
lizmat weekly: blogs.perl.org/users/damian_conway...lives.html 14:17
notable6 lizmat, Noted! (weekly)
librasteve posted to HN news.ycombinator.com/item?id=48775407, adding your comment in the next few minutes will count towards getting on the HN front page 14:22
lizmat +ed
librasteve (and then coleman and the other infra guys can pick up the pieces) 14:23
lizmat refrains from commenting, as that will probably cause downvotes 14:24
apogee_ntv Commented and upvoted 14:39
14:39 hwj joined 15:00 acidsys left 15:32 hwj left 15:37 librasteve_ left 15:51 acidsys joined 16:08 vrurg_ joined 16:11 vrurg left
SmokeMachine It seems the TRF registration form is not accepting special chars on name. (I found it out because my name is Fernando Corrêa de Oliveira… when I removed the ^ from my name it worked…) 16:15
[Coke] .tell librasteve 12:15 < SmokeMachine> It seems the TRF registration form is not accepting special chars on name. 16:51
tellable6 [Coke], I'll pass your message to librasteve
16:56 wayland76 joined 16:57 wayland left 17:18 apogee_ntv left 17:22 Pixi` joined 17:25 Pixi left 17:27 Pixi joined 17:29 Pixi` left 17:41 apogee_ntv joined 17:47 Sgeo joined, rmv left 19:08 guifa joined 19:12 guifa left 19:29 rmv joined, rmv left, rmv joined 19:47 vrurg joined 19:50 vrurg_ left
librasteve SmokeMachine: thanks for the feedback. For now, Air::Form validation is built on positive parsing of ASCII chars ... why? Well I do not have enough experience / tuits to make a full unicode round trip - as you know HARC stack is HTMX, Air, Red, Cro, so the desig intent is for all form fields to have validation and drop directly into a Red DB table - and I have no clue what should / should not be permitted in semantic table 21:23
fields. So I erred on the same/simple side.
tellable6 2026-07-03T16:51:33Z #raku <[Coke]> librasteve 12:15 < SmokeMachine> It seems the TRF registration form is not accepting special chars on name.
librasteve I suppose the ultimate would be accept any unicode name (eg Kanji, Tamil and so on) and to be thoughtful about mapping name to DB fields so that it can never be a key or sorted. 21:25
Before we get to that the next step is to move Air up to CRUD capabilities (as a simply blogging framework) and that needs Authentication and that needs user Registration (ie DB) and that isn't implemented in Cromponent ;-( 21:27
Any and all help to add tuits to the Air project to improve and accelerate would be gratefully received ;-)
Oh and I expect lizmat is going to want Air::Plugin::Commerce / Air::Plugin::Donate so that is higher on the list than blogging. 21:28
23:24 gcd left 23:31 gcd joined 23:57 rmv left

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!