Welcome to the main channel on the development of MoarVM, a virtual machine for NQP and Rakudo (moarvm.org). This channel is being logged for historical purposes.
Set by lizmat on 24 May 2021.
Nicholas good *able6, #moarvm 09:31
lizmat good moarning, Nicholas 09:36
jnthnwrthngtn moarning o/ 09:40
Nicholas \o
[Coke] ~~ 12:49
Nicholas +++ATH0 12:50
[Coke] ATA 12:57
I think we just hung up on each other.
[Coke] I come from the days of 300 baud but never spoke hayes. 12:58
Nicholas I believe I am later than that. But also I only know that one from later folklaw 12:59
[Coke] I had a 300 baud for my C64, if I remember the timeline accurately. 13:00
OH. I am very likely misremembering here, looking at dates. WTH, Coleda... 13:01
ah, no it's right, it's just that 300 baud was old tech even when I touched it. whew. 13:14
MasterDuke: I edited your comment on github (rude, sorry!) to reflect my github userid as opposed to my irc nick. 13:27
vrurg [Coke]: You still remember the commands... 14:57
japhb vrurg: Pretty much seared into the brain of anyone who had to debug a modem connection. 15:06
vrurg japhb: I have only fragments left. Though back in 90s it was possible to come up with init strings for 3-4 different modem brands. FIDONet times... :) 15:09
japhb Got me beat then. :-) 15:18
[Coke] vrurg, no, I had to google it. :) 15:20
one of my nerdiest moments in junior high was buying a numeric keypad input device for the C64 so I could enter in the assembly code printed COMPUTE! magazine to get "apps". (each line would have a checksum, and the programs would be pages long.) 15:22
vrurg [Coke]: I think this is the kind of thing basically everybody from that times got through – I mean, those printed codes. :) Ok, it's time to close the old farts club. Or I start recalling DEC assembler and why is it better than Z80/x86. :D 15:37
nine Repossession is just the worst! 15:51
I understand what it does. I understand completely why it has to do it and I also don't see another way of achieving the same than the one we have. It also costs a lot. Right now it's the major reason why we can't have concurrent module loading. 15:52
vrurg I wish somebody had time to have an article or a post about repossession. Don't think I ever understood it well. 15:56
MasterDuke [Coke]: cool, np 17:22
MasterDuke [Coke]: have you tried yet with a newer rakudo? if you're still seeing the problem i can merge my two open PRs and see if those help 20:24
[Coke] will try to do that this afternoon - wasn't it a newer moar? 20:29
MasterDuke well yeah, but i thought you said you were using rakudo 2022.02 20:30
and rakudo at current HEAD will pick up the newer moarvm 20:31
[Coke] no, using rakudo latest but there wasn't a moar bump at the time, so it was using moar 2022.02 21:09
MasterDuke ah 21:10
[Coke] gist.github.com/coke/79dc5c6601b8e...589e4aba6b 21:11
Actually, sitting here now I can try with latest commits. 21:13
MasterDuke i've just started my shift keeping black mesa secure, but assuming no scientific experiment goes awry and an alien invasions happens, i can have a look if you get any new results 21:21
[Coke] (doing a build with rakudo-latest and mimalloc)
heh 21:22
I am reminded that I basically stalled out on Demon Souls and need to put in some hours collecting them.
This is MoarVM version 2022.02-5-gec9fcaae2 built with JIT support 21:25
is that recent enough?
MasterDuke yep 21:27
[Coke] nope 21:29
rakudo of v2022.02-78-gd53b3ee83c, using mimalloc; same failure mode I was seeing earlier 21:30
anything else before I rebuild the mimalloc verison?
er, the *non*...
MasterDuke any idea what the code is doing when the invalid free happens? 21:32
[Coke] nope 21:54
I didn't golf it
(because every time I tried, stuff worked)
MasterDuke is there any nativecall? 21:56
[Coke] will try again.
not directly; using File::Find; JSON::Tiny; HTTP::UserAgent, don't think there's any there. 21:57
Geth MoarVM: a78776ea4c | (Daniel Green)++ | src/io/fileops.c
The result of a getenv() call should not be freed

  "The string pointed to by the return value of getenv() may be statically
allocated, and can be modified by a subsequent call to getenv(), putenv(3), setenv(3), or unsetenv(3)." - man 3 getenv
MoarVM: 2b96d18281 | MasterDuke17++ (committed using GitHub Web editor) | src/io/fileops.c
Merge pull request #1674 from MasterDuke17/do_not_free_results_of_getenv
MasterDuke then mind pulling moarvm and trying with ^^^? 21:58
[Coke] ok. just building/installing moarvm head and will try... 22:03
no change 22:04
is there an env var to force more GC pressure?
MasterDuke not an env var, but you can edit github.com/MoarVM/MoarVM/blob/mast...debug.h#L7 and rebuild 22:05
i did hope that pr might have fixed it. if you're using File::Find i figured maybe a `.e` was happening somewhere and causing problems 22:12
[Coke] down to 181 lines.. 22:14
heh. add debug output - nothing is printed. remove the code that follows the non-printed debug output: it's now printed. 22:30
MasterDuke computers are fun 22:33
[Coke] 139 lines and it seems to need a run call, a rakumod load (mine) that makes an http agent call. 22:35
MasterDuke afk for a bit, might not be back tonight 22:45