| Geth | MoarVM/debugserver_fixes: 848d3c238a | (Timo Paulssen)++ | src/debug/debugserver.c metadata & positionals for MVMCapture |
00:12 | |
| ab5tract | Ironically enough (given how much crap autotools has been getting), this longjmp scenario is a perfect example of where autoconf comes in handy | 01:21 | |
| timo: I think that the fallout from breaking QEMU would probably be just the tip of an iceberg if Windows were to remove this | 01:23 | ||
| timo | OTOH the qemu people surely have people who are sufficiently knowledgeable to deal with such a situation, and their user base is probably big enough that it would be noticed very quickly even in like preview builds of windows or something | 01:35 | |
| ab5tract | It can stand in as a bellwether for sure | 01:38 | |
| But also: if QEMU uses it, I bet there’s a significant pool of “dark code” that relies on it too, and MS is pretty consistent about backwards compat | 01:40 | ||
| timo | that's a soothing thought :) | 01:45 | |
|
02:12
Voldenet left
02:15
Voldenet joined
09:51
sena_kun joined
10:07
sena_kun left
15:07
[Coke] joined
|
|||
| [Coke] | non reproducible, but: gist.github.com/coke/20db0ddb89bc0...37654e0404 | 16:44 | |
| [0] >raku(26996,0x206e60f40) malloc: Heap corruption detected, free list is damaged at 0x6000020c0240 | |||
| I had just installed Terminal::LineEditor, then ran raku to open the REPL and ran 'say 3' | |||
| lizmat | hmmm could it be that Terminal::LineEditor needed to be re-precompiled ? | 16:50 | |
| and he the fact that it doesn't need to be any subsequent time, makes it unreproducible? | 16:51 | ||
| fwiw, I suspect we still have some race conditions in the "module installed but needs to be re-precompiled" path | 16:52 | ||
| [Coke] | I did a zef info before I installed it, it wasn't there. | 17:16 | |
| must have updated raku on this box at some point vaguely recently | |||
| lizmat | thing is that installing a module does NOT guarantee that it doesn't need to be re-precompiled for the $*REPO that you're working with | 17:17 | |
| the first time you use it, it will likely be re-precompiled again | |||
| and I suspect we still have a bug inbetween saving the bytecode, and then running it | 17:18 | ||
| [Coke] | ah | 17:36 | |
| lizmat | I wonder whether it is some premature optimization though: we just wrote the stuff from memory to the bytecode file, why not execute it from memory immediately as well | 17:38 | |
| and then have something gone awry just then | |||
|
21:04
sena_kun joined
21:16
sugarbeet left
21:17
sugarbeet joined
23:07
sena_kun left
|
|||