libera/#shirakumo - IRC Chatlog

10:03:46 karlosz shinmera: i think to fix this alignment issue we just need to change editcore to align to 4096

10:04:12 karlosz it kinda sucks that on linux we have 65536 as the page size so that's different

10:04:46 karlosz anyway, i can try like in 3 hours or so if you have the workstation up

10:33:50 Colleen <shinmera> karlosz: should be up now.

11:47:34 Colleen <shinmera> Also updated the trial-nx repo to include the sbcl-nx repo (nx branch) now.

12:28:42 Colleen <shinmera> And added a script to manage the subtrees more easily.

13:27:56 karlosz okay cool

13:30:42 karlosz hmmm, it looks like the page size on the armbian is actually 4096

13:30:44 karlosz weird

13:30:48 karlosz that makes things a bit easier though

13:31:14 karlosz i got alignment issues when i tried to make the page size 4kb on the gccfarm

13:37:15 karlosz okay, well i got the new binary now

13:37:25 karlosz only thing is that it dies in the same place

13:37:33 karlosz i just noticed the error is unrelated to a core not existing

13:37:37 karlosz this is the error i actually get

13:37:47 karlosz Error: nn::fs::GetEntryType() failed because the mount name was invalid.

13:38:11 Colleen <shinmera> Why's it trying to access the file system at all?

13:39:52 karlosz because even for shrinkwrapped cores it tries to open a file on disk

13:39:58 karlosz which it doesn't need to

13:40:01 karlosz it's just how the code is

13:40:04 karlosz maybe i can fix that

13:40:08 Colleen <shinmera> Ah.

13:47:15 karlosz hm, that's kinda complicated though

13:47:31 karlosz the code really relies on being able to have a file handle to its own executable

13:47:35 karlosz the question is why doesn't it work?

13:47:44 karlosz i see you stubbed in /app.nsop

13:47:57 karlosz app.nso rather

13:48:06 Colleen <shinmera> That was complete guesswork. I don't actually know how the filesystem works yet

13:48:17 Colleen <shinmera> It's very possible the application does *not* actually get to see its own executable.

13:48:32 Colleen <shinmera> And instead only gets to see stuff in the data/ dir, which also might need further setup.

13:48:49 Colleen <shinmera> What does it need the file handle for?

13:49:58 karlosz sbcl wants to know the position of the data section so it can do it's own load-time relocations in the data section (it's RW so it's fine)

13:50:03 karlosz the relocations are stored in the ELF

13:50:18 Colleen <shinmera> ah

13:52:10 karlosz i mean, it probably doesn't need to be done that way

13:52:20 karlosz it's just how it is done because the embedded core stuff already existed

13:52:25 Colleen <shinmera> Right

13:52:33 karlosz i think it would be most expedient to figure out if its possible

13:52:38 karlosz im going to look in the sdk

13:52:40 Colleen <shinmera> Another annoying complication

13:52:58 Colleen <shinmera> Given how locked down everything else is I'd be surprised if this was any different, tbh.

14:05:12 karlosz from what i can tell you can mount the ROM

14:05:23 karlosz but whether that includes the executable i don't know

14:06:17 Colleen <shinmera> can we easily call C++ code from the runtime, or do we need a separate shim library?

14:07:00 karlosz isn't it compiled by a C++ lcompiler?

14:07:20 Colleen <shinmera> pretty sure it is not

14:07:36 Colleen <shinmera> I remember trying to port things to work with a C++ compiler at one point but it's way too much

14:07:49 karlosz oh

14:07:50 Colleen <shinmera> the compiler is clang, not clang++

14:07:53 karlosz also um

14:07:59 karlosz can't we just put the exectuable itself in the resource folder

14:08:19 Colleen <shinmera> was thinking about that, yeah

14:08:53 Colleen <shinmera> that would work as long as it doesn't also try to gleam any other info from the self file pointer or something

14:09:17 karlosz no

14:09:25 karlosz ok this could work then

14:09:41 karlosz we just need to mount the rom in nnMain

14:09:55 karlosz and then we can just refer to it with a path like rom:/sbcl

14:10:09 Colleen <shinmera> also we might want to move this discussion to my private matrix server, cause uh, now we're entering some pretty dicey territory for nda

14:10:18 karlosz ok yes

14:10:35 Colleen <shinmera> you should be able to use your tynet account to log on https://matrix.tymoon.eu/

14:11:49 hayley ACTION should have stuck with the n-d-a nick

15:03:24 dthompso` ** NICK dthompson