Retro Rocket OS - A BASIC powered Operating System (started 2009!)

[willow ginkgo]

download and install the uefi roms for it

im not sure where you get them on macos, likely homebrew

or you likely just need to update that path

[willow ginkgo]

you might want to take the -vnc 0.0.0.0:2 out too

i use that for remote dev

[willow ginkgo]

adding some better error reporting

[willow ginkgo]

just spent 3 hours fixing the recursive descent parser madness that causes that error

got some cleanups to do on it tomorrow

but for now, sleep calls!

[willow ginkgo]

finally fixed unary operators so that it doesnt require a preparse step to 'represent' negative values without a negative number

[willow ginkgo]

https://www.youtube.com/shorts/7lY3L9fDFmc

[willow ginkgo]

• Now working on adding missing editor features; added search
• Fixed error where procedures inside functions break the expected return value
• Fixed issue where INSTR doesnt work right and can crash
• Fixed issue with move-to-first variable optimisation burying system variables like TRUE/FALSE and causing runtime errors in BASIC

[willow ginkgo]

finally squashed some deep interpreter bugs that were preventing search/replace working in the editor. nearly working now

[willow ginkgo]

decided to post on reddit
https://www.reddit.com/r/osdev/comments/1n2r6ti/end_to_end_os_dev_test_showing_graphics/

[willow ginkgo]

Added IPV4 loopback support (not needed it until now)

[willow ginkgo]

Last night i added:

New BASIC keywords

• OUTPORT
• OUTPORTW
• OUTPORTD
• UDPBIND
• UDPUNBIND
• UDPWRITE

New BASIC functions

• UDPREAD$
• UDPLASTSOURCEPORT
• UDPLASTIP$
• INPORT
• INPORTW
• INPORTD
• BITOR
• BITAND
• BITEOR
• BITNOT
• BITNOR
• BITNAND
• BITXNOR
• BITSHL
• BITSHR
• BITROL
• BITROR

New Programs

A PS2 mouse driver completely written in BASIC where you can query mouse state via a UDP socket on localhost. Needs a bit of testing.

[willow ginkgo]

well, it loads and it reads the mouse state. thats a start!

[willow ginkgo]

it works! im able to read mouse coordinates while the driver is loaded, by querying its port 🙂

UDPBIND "127.0.0.1", 9999
REPEAT
    UDPWRITE "127.0.0.1", 9999, 14501, "GET"
    PACKET$ = UDPREAD$(9999)
    IF PACKET$ <> "" THEN
        PRINT PACKET$
    ENDIF
UNTIL INKEY$ <> ""
UDPUNBIND "127.0.0.1", 9999

[timber dawn]

yay

[willow ginkgo]

mouse cursor being rendered from a separate program!

we now have working BASIC drivers 🙂

i might make a library for getting mouse x/y/b state so i dont need to duplicate this client code, going forward
https://github.com/brainboxdotcc/retro-rocket/blob/master/os/programs/mousetest.rrbasic

[willow ginkgo]

PEEK AND POKE, do you feel the DANGER yet?

[wild sedge]

do you identity map?

[willow ginkgo]

yup

retro rocket is: you break it, you keep it
aka mess around and find out

made a user library to interact with the mouse task

https://github.com/brainboxdotcc/retro-rocket/wiki/mouse

and also a page documenting tasks

https://github.com/brainboxdotcc/retro-rocket/wiki/Writing-Tasks

[timber dawn]

very epics

[upper locust]

what exactly is the difference between keywords and functions?

i assume just the syntax to use them?

[true ridge]

is there an option for command aliases some people might want to use like OUTB

or OUTW

[willow ginkgo]

aliases? nope, it wouldn't be hard to add them though.

just extra entries in an array

happy to accept a pr for it, see src/basic/function.c near the top

here's a list of all the new functions I added

• BITOR
• BITAND
• BITNOT
• BITEOR
• BITNAND
• BITNOR
• BITXNOR
• BITSHL
• BITSHR
• BITROL
• BITROR
• PEEK
• PEEKW
• PEEKD
• PEEKQ
• POKE
• POKEW
• POKED
• POKEQ
• INPORT
• OUTPORT
• INPORTW
• OUTPORTW
• INPORTD
• OUTPORTD
• INPORTQ
• OUTPORTQ

[true ridge]

ok fair

[true ridge]

that's a crap ton

nice

[willow ginkgo]

yeah needed a bunch of them for systems programming

I forgot about the udp ones lol there's even more

[true ridge]

oof

[willow ginkgo]

there, now they're added to the docs

[true ridge]

congrats

[willow ginkgo]

im increasing the LAPIC timer from 100hz to 1000hz

various housekeeping tasks run on it that are improved by a higher resolution, such as fragmented packet reassembly

[digital leaf]

you shouldnt have too many threads running at once, that will compromise responsiveness

you should have support for blocking waits in the kernel if you don't already

(like the thread doesnt get scheduled at all until some event happens)

[willow ginkgo]

yes, this is already a thing

right now, a BASIC process has a way to register a lightweight callback that is just expected to say when the thread can be released from sleep state

this isnt part of the timer system

so right now there are 3 processes:

at the time this snapshot of the process state was given to BASIC two processes were waiting on other processes

init was waiting on rocketsh and rocketsh was waiting on proclist

the waiting processes basically dont get scheduled, only their wakeup callback gets scheduled

its kind of hard to describe, because what im describing is an interpreter driven system, not native threads

just massively improved performance by moving the variables into 3 hashmaps instead of 3 linked lists

the linked list stuff was about 15 years old lol

[timber dawn]

you should make it so retro rocket can run executable files (like elf?) AND basic code

[willow ginkgo]

nope not happening, sorry

thats a rabbit hole i dont want the rabbit from

would mean a complete rewrite, to handle native multi tasking, which then means ring 3, because while i can trust my interprter at ring 0 i cant trust user binaries

which then means paging, vmm, and all the other rammel

goes from keep it simple into just what everyone else is doing

[digital leaf]

@willow ginkgo you might like this

https://github.com/microsoft/basic-m6502

(I know that there were disassemblies of MS BASIC before but this is the real deal)

[willow ginkgo]

i think ive seen it

[ebon iron]

lol

[willow ginkgo]

i had the disassembly of bbc basic once, thats very well documented and very clean

[ebon iron]

I'm curious whether Microsoft BASIC interpreted their BASIC files directly (they weren't plain ASCII files, but tokenized)

or if they ground it down further first

[willow ginkgo]

most BASICs back then were tokenized as you entered each line at the REPL

[ebon iron]

yeah

[willow ginkgo]

then strored in ram or on disk tokenized

i skip that step in my basic and just keep it all ASCII, its more portable

these days if i want to read a BBC BASIC program i need a viewer or converter to convert it back to ascii 🙁

[ebon iron]

yeah

it's annoying

same with my GW-BASIC programs

[willow ginkgo]

i think ive covered most of the scalability annoyances of my basic interpterer now

linked lists for var lists was ... ass

[digital leaf]

you might've seen disassemblies of it as this was published a couple hrs ago

[digital leaf]

💀

[willow ginkgo]

im currently working on properly accounting all bytes that are in use per basic program in proclist

which means moving absolutely everything about the context into the context buddy allocator

problem is stb_image used for sprites

it has no way to pass a context

you override its allocators via #define 🙁

#define STBI_NO_STDIO
#define STB_IMAGE_IMPLEMENTATION
#define STBI_NO_SIMD
#define STBI_NO_HDR
#define STBI_NO_THREAD_LOCALS 
#define STBI_MALLOC(sz)           kmalloc(sz)
#define STBI_REALLOC(p,newsz)     krealloc(p,newsz)
#define STBI_FREE(p)              kfree(p)
#include <stb_image.h>

gross 🤮

so no way to associate allocations with some owning context

[heavy linden]

Basic JIT when?

[glad hemlock]

You allow direct port io within basic?

I thought that was your userland or equivalent

[willow ginkgo]

yes

direct port io and direct identity mapped access, just like the BASIC of old

peek it, poke it, you break it you keep it

i may put some settings in at some point that restrict which BASIC programs have port io and direct peek/poke and which dont, purely a flag you toggle

its not a multi user system, doesnt intend to be, so its all purely for experimenting.... want to poke at something see what happens? retro rocket got your back

[glad hemlock]

Yeah that would probably be a good idea

But yeah if you’re trying to make basic like back in the day you need that

Cool that you can make drivers within your own system though

Do you plan on doing privilege levels?

I know it’s not a multiuser system but

[willow ginkgo]

depends what you mean

like normal user, administrator?

maybe, not sure

[glad hemlock]

Yeah kinda like Unix

[glad hemlock]

Maybe certain programs can have privilege level

Like programs that interact with the hardware could require some sort of password or acknowledgement from the user that the program could very well nuke the computer

Like If a program uses port io, make it prompt the user or require something like sudo to make it run due to it being possible to crash

Idk

[true ridge]

why would BASIC have a user system

at most just have a popup saying "program might do something funny to system plz beware"

at best just have nothing

[glad hemlock]

That’s what I mean

[willow ginkgo]

not sure

the itent isnt to restrict stuff much

there are some obvious safeguards like you cant PEEK or POKE ram that doesnt exist (including ram holes)

dont want an obvious pagefault

perhaps some system where the first programs that start, start with ability to do everything

and, programs can disable keywords and functions for child programs, a child program cant then re-enable

so it works like a chroot or jail

so if you want to restrict it, you'd edit init or whatever

[true ridge]

cool

[glad hemlock]

Yeah I was thinking about something like chroot

[willow ginkgo]

i think if i just provide that, a way for a program to restrict what its child can do

that will be enough

if someone then wants to make permission systems, login prompts, user levels it can all be done thru basic

if i make it so you can also restrict what parts of the filesystem the child can see

[true ridge]

retro rocket microkernel?

[willow ginkgo]

is it a microkernel if everything is ring 0 anyway? lol

[glad hemlock]

Probably not

But

You can call it whatever you want

And as long as you believe in it

It’s the truth

[true ridge]

your basic interpreter might be the "user space"

[willow ginkgo]

it essentially is, that's the idea, any restrictions are part of the interpreter, no context switches

[true ridge]

understandable

[willow ginkgo]

really proud of this.... got the LINKER to sort an array by name of the symbol within it

        . = ALIGN(8);
        __start_kw_array = .;
        KEEP(*(SORT_BY_NAME(.kw.*)))
        __stop_kw_array = .;

basically takes a whole load of keyword name sections like .kw.ELSE, .kw.FOR
then sorts the section at link time so when you reference it, you get an array of integers, sorted by the name of the keyword

which is needed for fast search of tokens in the tokenizer

i was having to manually maintain two different lists, now i dont

[true ridge]

why would you do that with the LINKER first of all

is there a point for this?

[true ridge]

oh

nevermind

cool tho congratulations on your linker sorter

i need to read everything before responding :/

[willow ginkgo]

lol

just rush right on in there 😄

im currently trying to build an elf64 module loader

so i can move stuff out of the kernel core

[true ridge]

yoooooooooooo

actually i wanna try the os how do i boot it (how 2 compile)

[willow ginkgo]

git clone, make sure you got gcc, mtools, php, xorriso, nasm... i think thats it

oh and cmake

its a cmake project

or if you cba to build it.... there are github actions that build isos

along with a run.sh

[true ridge]

based

[true ridge]

have these

is it a cross compiler gcc or just gcc

[willow ginkgo]

just gcc

x86_64

[true ridge]

k

[willow ginkgo]

i did have a freestanding cross compiler years ago

but i lost it, didnt bother to reinstall another, and everything just worked fine so

[true ridge]

https://tenor.com/view/i-don't-know-idk-idk-about-that-gif-7336250873949261946

[willow ginkgo]

but the main reason i had it was i used to build on a 32 bit netbook

lol

[true ridge]

the superiority of the intel atom 🔥

[willow ginkgo]

lmao you know it

asus eeeeeeeeepc

[true ridge]

eeeeeeeeeeee peee ceee

i have a netbook its some shitty compaq one

[willow ginkgo]

mine still boots, got ubuntu 16 on it iirc or maybe 14

i dont use it though

only 2gb ram

[true ridge]

once i got literally any other laptop from that era the core 2 duo's felt like actual bliss

[true ridge]

nice

[true ridge]

i had 1 😭

[willow ginkgo]

the ./run.sh the build generates, you might need to adjust

i have it set to host the screen via vnc

you probably dont want that 😄

[true ridge]

i forgor you're using vnc

im probably just gonna use qemu

or virtualbox

[willow ginkgo]

yeah its qemu

but the qemu run command set vnc for the screen/keyboard etc

just take the -vnc flag out

[true ridge]

wait i remembered that the core 2 duo laptop i have matches the system requirements for this i might run it there

[willow ginkgo]

qemu-system-x86_64  -machine q35,accel=kvm  -s  -monitor stdio  -cpu host  --enable-kvm  -smp 8  -m 4096  -drive id=disk,file=../../harddisk0,format=raw,if=none  -device ahci,id=ahci  -device ide-hd,drive=disk,bus=ahci.0  -drive file=rr.iso,media=cdrom,if=none,id=sata-cdrom  -device ide-cd,drive=sata-cdrom,bus=ahci.1  -no-shutdown  -boot d  -vnc 0.0.0.0:2  -debugcon file:debug.log -netdev user,id=netuser,hostfwd=udp::2000-:2000  -object filter-dump,id=dump,netdev=netuser,file=dump.dat  -device e1000,netdev=netuser

[true ridge]

k

[willow ginkgo]

its a bit of a monster lol

[true ridge]

oh god this config

[willow ginkgo]

this is why the cmake outputs the ./run.sh for you, ready to go

[true ridge]

oh cool

[willow ginkgo]

imagine having to reconfigure it each time

[true ridge]

[willow ginkgo]

it makes some assumptions: "../../harddisk0" is a 1gb (or however big) empty raw disk image

theres a few hairs, because i havent had many others play with it for a while

[true ridge]

does retro rocket (ima just call it rr) format the disk image?

[willow ginkgo]

only if you install it via the install option from the boot menu

[true ridge]

understandable

is the installer also a basic program

[willow ginkgo]

after that you can actually boot the hard disk image

no, it isnt

[true ridge]

dangit

[willow ginkgo]

i didnt go that far

[true ridge]

lame /j

[willow ginkgo]

it lacks a couple of features i wasnt sure i wanted to give... like write raw sectors. i'd have to reinvent half the filesystem in basic

[true ridge]

i mean understandable

security is gonna be done for if some program just overwrote the hard drive

[willow ginkgo]

read raw sectors might be a cool BASIC feature though, imagine a disk sector hex dumper

[true ridge]

it's not even supposed to be secure but it would be quite a mindfuck to reinstall this for 554th time

[willow ginkgo]

lol

[true ridge]

would be cool

[willow ginkgo]

i reinstall mine like every week

every time theres new kernel

because i dont have an upgrade path yet

i mean it isnt difficult to do, just dont reformat the rfs

just havent gotten around to it

[true ridge]

id probably just have the makefile somehow install the kernel into a disk image because thats what i usually do

[willow ginkgo]

it does

[true ridge]

reasonable

are you planning to make an upgrade path (eventually)?

[willow ginkgo]

when you do a hard disk install it flattens the entire drive, then makes two GPT partitions:

1. EFI recovery, FAT32 68mb, contains kernel and limine
2. the rest of the disk RFS, contains the BASIC programs, /programs/init, rocketsh, /system.

the EFI recovery is written out directly from a .img on the cdrom

made by the cmake

[true ridge]

custom file system :0

[willow ginkgo]

yup!

[true ridge]

peak memcpy (memcpy doesnt even copy to the hard drive but still) /j

[willow ginkgo]

its streamed

[true ridge]

yooo nice

[true ridge]

cool

[willow ginkgo]

once im done playing about with random stuff like kernel module support... i was improving the editor

after improving the editor, was going to make a package manager, in basic

that fetches basic programs from a http repo

[true ridge]

do you have basic syntax highlighting actually?

[willow ginkgo]

no, thats on my todo next in the editor feature list

[true ridge]

oh

[willow ginkgo]

for BASIC syntax highlighting i currently work in clion

[true ridge]

i cant have colored text while doing my 10 PRINT "Hello World" 20 GOTO 10 :(

[willow ginkgo]

oh, line numbers

use them if you like

[true ridge]

nice

[willow ginkgo]

but theyre discouraged

[true ridge]

oh

labels?

[willow ginkgo]

nah

i didnt want to encourage use of goto

we got PROCs and FNs and LIBRARY

[true ridge]

what should be used instead im not really familiar with basic

[willow ginkgo]

PROC is a procedure, like a function but never returns a value, can take params

[true ridge]

void

nice

[willow ginkgo]

PROCfoo("hello")

DEF PROCfoo(word$)
    PRINT word$; " world"
ENDPROC

FN is a function, can take params but instead of ENDPROC, ends with = and a value

B = FNcrap(7,8)

DEF FNcrap(B, C)
    D = B + C
=D

[true ridge]

=D

[willow ginkgo]

a LIBRARY is a collection of PROC and FN in a separate file, that you can load with the LIBRARY keyword, if you use LIBRARY you shouldnt line numbers

LIBRARY works like #include except it appends to the end, not inplace

[true ridge]

how does that work??? (does it just add the library functions at the end of your file?)

[willow ginkgo]

and if you define a PROC in a library with the same name as the library, it will be auto called on load like a constructor

[willow ginkgo]

when you call LIBRARY the basic program is krealloc'd, to make space on the end, then its internally renumbered, reparsed, and the library appended on the end

after that the whole combined program is re-scanned for procedures and functions (you dont need forward declarations)

and if it exists, the constructor is ran

[true ridge]

oh

[willow ginkgo]

it then sets the program ptr to continue after the lib statement

[true ridge]

fair enough

[true ridge]

understandable

[willow ginkgo]

two ways to start a program: from the command line, by name (it looks in /programs) or by the CHAIN command

CHAIN "/program/name/goes/here"

if you start a program by name from the shell, command line args are passed in, if you CHAIN it they arent

[true ridge]

ok wait sorry to interrupt you but got an error after getting the makefile from cmake

make[2]: *** No rule to make target '~/Downloads/rr/retro-rocket-master/limine/limine.sys', needed by 'iso/limine.sys'.  Stop.

i need to just put limine somewhere

[willow ginkgo]

oh, its a specific version, git submodule iirc

brain@neuron:~/retrorocket/cmake-build-debug$ git submodule
245c7c94c20eac22730ef89035967f78b77bf405 ../doxygen-awesome-css (v2.2.0-2-g245c7c9)
c93da495943c567340b7e128ae5f4fc5eaf17f43 ../limine (v4.20231024.eol-binary)

if you did git clone --recursive it would fetch it

[true ridge]

ah

[willow ginkgo]

cant remember the command to do it after the fact

[true ridge]

can i use a different version or am i gonna get smitten by zeus himself or something like that

[willow ginkgo]

youll get struck down by lightning

limine changed their config format completely

[true ridge]

JUDGEMENT'd

[willow ginkgo]

i cba to update to the new format

it would give me some advantages though like better customisation of the boot menu

[true ridge]

fair

hobby os's being evil asf for no particular reason as usual

[willow ginkgo]

lol

aaaaaaaaaaaaaaaaaaaaaaaaaah

they archived the github

i think i might just make the binaries part of my repo

i really dont want to deal with codeberg

i dont chase trends

[true ridge]

fatal: not a git repository (or any of the parent directories): .git
bash: syntax error near unexpected token `('
bash: syntax error near unexpected token `('

[willow ginkgo]

lol

where are you getting this from

[true ridge]

this command

[willow ginkgo]

oh

you dont need this submodule anyway

its only for docs gen

[true ridge]

why did they even move there

[willow ginkgo]

idk

drama

apparently mintsuki is banned from gh

[true ridge]

im probably blind as shit i dont see anything

[true ridge]

yikes

[willow ginkgo]

you know you dont have to build your own rr?

[true ridge]

no :o

i didnt see any releases on the github so i just assumed "ok i have to compile it then"

[willow ginkgo]

ugh master is broken anyway

why didnt i spot this

[ 71%] Generating iso/system/timezones/Pacific/Palau
make[2]: *** No rule to make target 'CMakeFiles/mod_test.dir/modules/test/test.c.o', needed by 'iso/system/modules/test.ko'.  Stop.
[ 72%] Generating iso/system/timezones/Pacific/Pitcairn
make[1]: *** [CMakeFiles/Makefile2:11616: CMakeFiles/module_test.dir/all] Error 2
make[1]: *** Waiting for unfinished jobs....
[ 72%] Generating iso/system/timezones/Pacific/Tahiti

it was building fine

[true ridge]

https://tenor.com/view/jet-bean-killer-bean-ragdoll-ragdoll-fall-gif-23762382

rip

[willow ginkgo]

i'll fix this then i probably should slep

its like 3:45am

[true ridge]

[willow ginkgo]

https://tenor.com/view/dead-my-honest-reaction-damn-skull-skeleton-gif-15885937969029652188

[true ridge]

https://tenor.com/view/funny-skeleton-man2-cryos-funny-skeleton-man-gif-27695831

tell me when the thing gets fixed

ping idk

[willow ginkgo]

there, its fixed @true ridge

wasnt complex to fix

but the reason i noticed was i checked gh actions

if youre strugglin to build just grab iso and shell script: https://github.com/brainboxdotcc/retro-rocket/actions/runs/17482255429

scroll down

[true ridge]

k

[willow ginkgo]

hah installer is broken

i'll fix that next, easy fix i think

[true ridge]

i made it run

nice

resolution is insanely big my screen is too small lmao

ball works

TEXTURED CUBE

[willow ginkgo]

installer fixed

[true ridge]

YOOOOOOO

[true ridge]

nice

[willow ginkgo]

you found it then

[true ridge]

the mousetest doesnt mouse

[willow ginkgo]

you must load the driver first

[true ridge]

oh

[willow ginkgo]

task drivers/ps2mouse

then run mousetest

it isnt auto loaded as its part of my longer term thing to maybe make a gui... in basic, and it will auto load the task

[true ridge]

my mouse mouses

[true ridge]

BASIC init system

[willow ginkgo]

🐭

btw list /programs/init

there actually is an init program, written in basic, its the first thing that runs that launches rocketsh

[true ridge]

autoexec

[willow ginkgo]

which is also in basic

[true ridge]

this is truly basic

gradient sine wave

[willow ginkgo]

unbasic basic

[true ridge]

how to clear screen i have artifacts from the demos

[willow ginkgo]

CLS

(in caps, its a basic keyword)

[true ridge]

it's case sensitive :(

[willow ginkgo]

the shell prompt is actually a repl

[true ridge]

cool

[willow ginkgo]

try PRINT 1 + 3 or PRINT "HELLO WORLD"

yeah BASIC is case sensitive btw

some dialects got rid of that

[true ridge]

[willow ginkgo]

[true ridge]

photoshop

[willow ginkgo]

figured lol

[true ridge]

a

didnt know that

[true ridge]

ok it does print 4 if in actuality

[willow ginkgo]

try this

INPUT VAR$

then type some shit on the blank line

then do PRINT VAR$

[true ridge]

ignore the error i tried something stupid

cool

can i put text before the input box

wait the test demo did that

[willow ginkgo]

sure, in an actual program but not on the REPL

[true ridge]

fair

[willow ginkgo]

in a program you'd do:

PRINT "Prompt here";
INPUT VAR$
PRINT "You said "; VAR$; "!"

the ; on the end of the print line means no new line

you can separate items to PRINT with , or ;

[true ridge]

placing question marks doesnt work

[willow ginkgo]

; means put it exactly there, and , means use tabulation

you mean the ? key?

[true ridge]

k

[true ridge]

yes

[willow ginkgo]

hmm, probably a bad keymap

[true ridge]

maybe

prints a "/"

[willow ginkgo]

yeah can duplicate that

[true ridge]

mk

i managed to break the editor tool :'c

[willow ginkgo]

lol

what did you do

oh did you read the license agrement, said every bug you find you gotta submit a pr fixing it /j

[true ridge]

apparently it's using the "REPLACE" option

on an empty file i think

[willow ginkgo]

lol

[true ridge]

idk

it works on any file

[willow ginkgo]

never thought to test that

what does it do?

btw, if any basic program locks up, ctrl+esc should drop you back to shell prompt

its an interrupt key like sigint

that is unless the interpreter is ded

[true ridge]

so if i do the funny a few times it might give me its gonna give me an error on line 489

[true ridge]

GOD DAMN IT

[true ridge]

idk why i spammed ctrl+r but yeah idk it died

rip

[willow ginkgo]

lol

back to shell prompt?

[true ridge]

yes

shell works tho

[willow ginkgo]

well, not so bad, at least whole os didnt go

[true ridge]

breaking news: dumbass spams ctrl+r and destroys retro rocket

[willow ginkgo]

heres some commands to try
about, version, dir

[true ridge]

tried dir already

and cd

[willow ginkgo]

also, to go up a dir the command is just 'up'

rather than cd ..

[true ridge]

52034290302 color bitmaps :0

ignore the dying editor

(the dead)

[willow ginkgo]

haha, your screen is mess

[true ridge]

:P

but does it work on real hardware (dvd burning time MUAHAHAHAHAHAHA)

[willow ginkgo]

yes

at least:

cant guarantee anything else

[true ridge]

absolutely Machine Mine Work On

ima try it on a core 2 duo for shits and giggles

[willow ginkgo]

im running it rn downstairs in my cellar actually, testing a dhcp fix

there was horrible bug that cost me a whole damn weekend, not of OSdev but of network troubleshooting

[true ridge]

hobby os's all need a certain purpose in life

[willow ginkgo]

if you left it running, it would eat all the dhcp leases over 100 hours uptime

[true ridge]

networks are stupid for no reason at all

idk why

[willow ginkgo]

imagine whole family, 3 kids and wife complaining they cant get online

its been up since weds now and managing to hold onto the same ip

[true ridge]

lmao

[willow ginkgo]

can still ping it, must still be running

brain@neuron:~/retrorocket/cmake-build-debug$ ping 10.0.0.157
PING 10.0.0.157 (10.0.0.157) 56(84) bytes of data.
64 bytes from 10.0.0.157: icmp_seq=1 ttl=64 time=0.399 ms
64 bytes from 10.0.0.157: icmp_seq=2 ttl=64 time=0.149 ms
64 bytes from 10.0.0.157: icmp_seq=3 ttl=64 time=0.158 ms
64 bytes from 10.0.0.157: icmp_seq=4 ttl=64 time=0.214 ms
^C
--- 10.0.0.157 ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3073ms
rtt min/avg/max/mdev = 0.149/0.230/0.399/0.100 ms

[true ridge]

50 000 dollar bet on that it dies right NOW

[willow ginkgo]

lmao

probably could kill it if i tried

its probably vulnerable to all kinds of ip exploits

[true ridge]

hackers walking in thinking they got another windows pc

3 seconds after connecting they quickly reconsidered life choices

[willow ginkgo]

lol

hackers: nice it pings what is it
also hackers: wtf i this, whats a BASIC?

[true ridge]

"what the fuck is a retro rocket"

[willow ginkgo]

"nice we hacked nasa"

[true ridge]

the cmos clock commited die on the core 2 duo laptop

the battery might be a safety hazard the icon is flashing red then blue

these second hand laptops go crazy 😭

[willow ginkgo]

lol

my laptop is also a refurb

[true ridge]

also vs2010 and windows 7 :0 (i remembered my funny ass directx 3d test project i made)

[true ridge]

nice

[willow ginkgo]

my laptop

[true ridge]

KUBUNTU USER YOOOOOO

also thinkpad

[willow ginkgo]

thinkpads rock

really repairable, durable, snappy

[true ridge]

i have one but its 32 bit so im sad i can only get windows xp on it

i got 7 on it but its too old to have drivers working

[willow ginkgo]

sadge

[true ridge]

its from like 2004

idk its a surprisingly good windows 98 machine

literally bought it because it ran windows 98 i wanted to try playing multiplayer dos games
-# with myself tho this is saddening asf :c

im gonna try writing the iso on a usb for shits and giggles

[willow ginkgo]

it wont boot

limine will load the kernel, you may get as far as mounting the root fs, then it'll die

because no usb support

[true ridge]

awh

[willow ginkgo]

if anything different happens lmk

[true ridge]

im gonna try it for the science tho

[willow ginkgo]

but youll probably get "failed to mount / as iso9660"

https://tenor.com/view/breakingbad-science-yeah-science-gif-5954775

[true ridge]

probs

[willow ginkgo]

damnit 4.30am im a bad adult

[true ridge]

limine stuck on flashing cursor

[willow ginkgo]

lmao

[true ridge]

rip

i lost the connection cable for my usb cdrom drive brb

[willow ginkgo]

mine has cable built into it

[true ridge]

apparently i was charging a calculator with it

dementia

[true ridge]

luckyn't

time to sacrifice a cd-rw that has a 50% chance of reading

what if a read fails actually

[willow ginkgo]

depends

in retro rocket you'll get an error

in limine idk

[true ridge]

ok it booted

i had to disable "intel rapid storage technology" or whatever tf its called

congratbulstations

textured cube runs fast

or at least like 60 fps

[willow ginkgo]

nice!

[willow ginkgo]

this is actually really good to see, first time someone's running it on real hardware except for me

[willow ginkgo]

this is normally some kind of ide emulation

dressed up as some kinda raid

[true ridge]

its some kind of sata extension iirc

but its a bit wonky

[willow ginkgo]

did you dare try to install it? 😛

[true ridge]

i dont want to overwrite my windows 7 install

i have a spare hard drive when i get home its getting sacrificed

its like 140 gb

[willow ginkgo]

can that laptop boot from uefi?

like with an EFI partition and GPT?

because i dont do old fashioned bios partitioning

[true ridge]

no

[willow ginkgo]

hmm, it prob wont boot then

[true ridge]

theres efi emulators iirc

[willow ginkgo]

but you CAN use the installer to make an rfs partition, then mount the rfs partition from the livecd

[true ridge]

didnt know this boots from efi tho either i forgor or something else

[willow ginkgo]

how old is it

[true ridge]

i think 2017

[willow ginkgo]

it should then

its not ancient

the machine i test on boots from efi and that is 2012

[true ridge]

its def uefi

[true ridge]

nice

[willow ginkgo]

should be fine

[true ridge]

its hard drive is borked tho and it only has nvme

[willow ginkgo]

i tried it in qemu on this work laptop... something weird happens in the basic intepreter

if i have time some point i'll try and figure it out

[true ridge]

weird

[willow ginkgo]

something about it doesnt look right though its its an old version

either that or it can boot to shell in 0.6 secs

[true ridge]

that could be the problem idk they changed something maybe

[willow ginkgo]

yeah theyre constantly changing the q35 emulation

to make it more compatible with windows and linux

[true ridge]

mainstream os's

[willow ginkgo]

boooo yeah

my end goal is to daily drive retro rocket, admittedly side by side with windows/linux on its own pc, not replacing it

[true ridge]

same with my os except its not the entire universe propped up by a basic interpreter

[willow ginkgo]

lol

you porting gnu userland?

[true ridge]

didnt get that far :/

[willow ginkgo]

wish someone would port bsd userland instead

just for shits and giggles

[true ridge]

i got a vmm recently its not very far at all

[willow ginkgo]

once ive got my module loader working, theres nothing stopping someone dynamically loading native code into the kernel at runtime

[true ridge]

are these utils or wdym by userland

[willow ginkgo]

like all the shell, utils, etc

instead of gnu, the ones from freebsd

tcsh, csh, bsd net tools etc

[true ridge]

tfw you took time to make a basic interpreter but then someone just uses linux apps with retro rocket (compatibility layer )

[willow ginkgo]

much of the code is a lot cleaner than gnu's tbh

lmao

i'd be curious to see that

[true ridge]

i never used freebsd (sad)

[willow ginkgo]

someone makes a driver that just co-opts the scheduler, starts a pre-emptive interrupt driven one and drops to ring 3.. lmao

and i'd be like wtf

[true ridge]

lmfao

[willow ginkgo]

tbh i dont know if my module loader works yet

i havent tested it

[true ridge]

the utils are probably good but im not making a unix like (maybe they should still work)

[willow ginkgo]

i wrote it at 2am with assistance of caffeine, google, my old ELF64 code, and a bit of LLM

[true ridge]

make a module that prints something idk

[willow ginkgo]

i have one

i just havent tested it works

[true ridge]

oh

[willow ginkgo]

i was tempted to, but then we started chatting, and i realised that if i diverted away to test it im 50% sure it will break, and then i'd be up another 3 hours fixing it

[true ridge]

are modules entered through the limine boot argument thing

[willow ginkgo]

so i shelved that step for the weekend

no, they'll be loaded from the root fs

[true ridge]

sorry

[willow ginkgo]

i mean they could be loaded via limine modules, but that isnt my plan

[true ridge]

mac os <10 extension ptsd

[willow ginkgo]

that whole limine thing is a weird preboot environment i dont want to linger in

because on a hard disk its all inside a fat32 partition

and i hate fat32

my fat32 driver is spotty at best

[true ridge]

i dont get the luxury of escaping limine boot args i want a microkernel

[willow ginkgo]

i wouldnt trust it to relaibly write modules into the efi

[true ridge]

understandable

[willow ginkgo]

my plan is to have two things: 1) a basic keyword to load a kernel module and 2) a command line program that just calls it

so a program, like /programs/init can do KMODULE "/system/modules/test.ko"

or from rocketsh you can just do module /system/modules/test.ko

[true ridge]

whats the extension for basic software in the os actually
or is there just none

[willow ginkgo]

or event just module test

[true ridge]

nice

[willow ginkgo]

theres none

in the cmake environment before the iso's built they have the extension .rrbasic, but this is just so the syntax highlighter works

theyre renamed to have no extension in <CMAKEBUILD>/iso

im going to need some kinda heuristic to know if to syntax highlight, just realised this, boo

[true ridge]

if in rr itself you save smth with .rrbasic (nobody except psychotic dumbasses will do that but still) will it still run

[willow ginkgo]

sure

extensions arent an important thing

so long as you include that exension as part of the name

[true ridge]

a

[willow ginkgo]

it wont try to append it

[true ridge]

if you do syntax highlighting you can probably detect if its compatible with your basic dialect

so its not just a text file getting highlighted or whatever

[willow ginkgo]

it would probably be as simple as "does the first non blank line contain at least one BASIC keyword or a variable assignment"

[true ridge]

fair

[true ridge]

yeah maybe smth like that

scan the whole file and make sure its basic compatible

[willow ginkgo]

lmao

i dont even do that when running one

[true ridge]

that i did notice

also basic interpreters usually just crash if they find something stupid anyways

[willow ginkgo]

yup

i took time to give actual useful error messages

theres a minor bug though if the line contains tabs, the ^ that is supposed to point at the error character doesnt tab with it

[true ridge]

i need to learn the language tho

[true ridge]

rip

[willow ginkgo]

did you check the wiki

thing is you cant actually make your own programs unless you install because theres nowhere to save on the cd

unless, you drop your own .rrbasic into os/programs in cmake

and then do make, and burn your own custom iso

[true ridge]

uhhhh yeah for the shell commands

[willow ginkgo]

tbh thats how i test rn, its kinda "long form" :S

[true ridge]

fair

[willow ginkgo]

https://github.com/brainboxdotcc/retro-rocket/wiki/Creating-BASIC-programs

[true ridge]

spare hard drives saving the universe again

[willow ginkgo]

time to get back to work, speak later

[true ridge]

cya

[willow ginkgo]

it works!!!

    module test;
    fs_directory_entry_t * f = fs_get_file_info("/system/modules/test.ko");
    if (f) {
        dprintf("*** MOD TEST: Got info on test.ko size %lu\n", f->size);
        void* module_content = kmalloc(f->size);
        if (fs_read_file(f, 0, f->size, module_content)) {
            dprintf("*** LOADING TEST MODULE test.ko ***\n");
            if (module_load_from_memory(module_content, f->size, &test)) {
                module_unload(&test);
            }
            dprintf("*** DONE WITH TEST MODULE test.ko ***\n");
        } else {
            dprintf("Couldnt read test.ko\n");
        }
    } else {
        dprintf("*** MOD TEST: Cant get info on test.ko\n");
    }

[1427]: *** MOD TEST: Got info on test.ko size 5192
[1428]: *** LOADING TEST MODULE test.ko ***
[1428]: test.ko: mod_init called!
[1429]: test.ko: mod_exit called!
[1429]: *** DONE WITH TEST MODULE test.ko ***

[true ridge]

LETS GOOO

rr as a linux loader when

[willow ginkgo]

lol

[true ridge]

are modules gonna work like programs in a sane regular os

it feels like theyre just drivers

[latent tendon]

modules are just any optional piece of code

They're not programs since they link to the kernel and exist in the same address space

[true ridge]

ohok

[willow ginkgo]

yesnt, they have full access to all symbols in the kernel, and are basically exactly a .so

nothing else special except some compiler/linker flags to stop them being position independent and stop them using 2gb relocs

so you could schedule a timer, and have it called on events

[willow ginkgo]

technically in retro rocket everything exists in the same address space 🙂

kernel modules in rr are placed in a kmalloc'ed aligned region big enough to hold all the PT_LOAD sections

which can be anywhere in the heap, hence the compile flag for mcmodel=large

was simply the easiest way to deal with it in a non-paged, identity mapped layout

[true ridge]

ohokcool

[willow ginkgo]

started moving my docs from github wiki into doxygen, gives me much more flexibility of layout

i got so many docs 🙁

[true ridge]

idk what doxygen is :'(

[true ridge]

that's a good thing especially if they're well written

any dumbass can just go and learn basic

[willow ginkgo]

https://retrorocket.dev

^ doxygen

my docs need a bit of a cleanup tbh

[true ridge]

where's the flashy visuals that make me feel like in a cartoon /j

where's the rocket owl did the rocket fail

[willow ginkgo]

lol

i havent had time to redo the graphics yet

those are just the github ones

[true ridge]

understandable

are they still gonna be in markdown or whatever you used on github?

is the github wiki gonna be maintained?

[willow ginkgo]

no

everything has been copied to the doxygen one

which is easily maintained by PRs

its a bunch of markdown files in the docpages dir

[true ridge]

sacrifices have to be made 😥

wrong message

[willow ginkgo]

the github one only lets me have like 1 level of pages

[true ridge]

whatever

[true ridge]

oh

lame

[willow ginkgo]

on gh wiki on the right, that huge sidebar

that is the best i can do, cant collapse or expand etc

[true ridge]

saddening

[willow ginkgo]

have now made e1000 and rtl8139 into modules! right now the kernel hard codes loading e1000 at the end of the startup process before starting the scheduler, but the next step later tonight will be to give a BASIC keyword to load a kernel module, and let /programs/init load the driver

[true ridge]

intel 8254x when?

[true ridge]

yay internet

[willow ginkgo]

thats what e1000 is

or rather a subset of

82540EM and 82541PI

82540EM is qemu

82541PI i have in my test machine

[true ridge]

:0

i have a thinkpad with a network card that series but it wont run RR it's 32 bit and ide only :(

(8254x series)

also can modules have their own keywords or smth

for basic

actually no wait

im stupid pardon me

[willow ginkgo]

there, now BASIC can load and unload kernel modules!

[true ridge]

this is a basic of all time

[willow ginkgo]

of all time!!!

[willow ginkgo]

can now configure network via a config file in /system/config/network.conf, the defaults are for dhcp on everything. once its installed you can edit this.

[plush estuary]

BOAT

[willow ginkgo]

been busy tonight! lots of bugs squished:

• Fixed hostname bug: stopped freeing string literal; now store hostname in static buffer.
• Corrected IRQ stub: pass real IRQ number to C handler (was always 0).
• Implemented proper shared-IRQ handling; call all handlers on shared line.
• AHCI ISR now acks/clears PxIS bits per port to prevent interrupt storms.
• e1000 init hardened: fully initialize TX/RX descriptor rings and indices.
• Zeroed TX buffers/descriptors so no stale bytes leak into frames.
• Programmed TX/RX descriptor base/len registers correctly and aligned buffers.
• Verified on HDD boot under QEMU: no bogus 0xFF frame

[true ridge]

it took you this long since you first programmed your idt to realise that the irq handler always returns 0

damn

[willow ginkgo]

it was one of the params in the irq handler function

[true ridge]

still though

[willow ginkgo]

void e1000_handler([[maybe_unused]] uint8_t isr, [[maybe_unused]] uint64_t error, [[maybe_unused]] uint64_t irq, void* opaque) {
    uint32_t status = e1000_read_command(REG_ICR);
    if (status & ICR_RXT0) {
        e1000_handle_receive();
    }
}

^ the value of irq was always 0

never noticed, as nothing actually uses it

purely informational

[true ridge]

oh

[willow ginkgo]

the assembler macro that built the jump table was just setting it to $0

[true ridge]

https://cdn.discordapp.com/attachments/1406827148068782171/1414358174277963860/flooshed.gif

a

[true ridge]

rip

[willow ginkgo]

all good now

ugh

what on earth reminna

[true ridge]

all good now

[willow ginkgo]

this isnt 'when your os goes crazy' this is when reminna goes crazy and cant take a simple screenshot in wayland lol

https://tenor.com/view/thanos-gif-25117042

[true ridge]

i almost just said you should post it there

[willow ginkgo]

im sick of seeing this tho

[true ridge]

wtfisthat

[willow ginkgo]

there, all good

[willow ginkgo]

wireshark

[true ridge]

welcome to the internet

[willow ginkgo]

if you ever integrate or write a tcp/ip stack, get used to spending hours in it

[true ridge]

oh

internet will destroy my sanity i assume :/

porting a web browser will be total ass as well probably

[willow ginkgo]

im contemplating making one

and making my life easy by

1. not implementing ssl yet
2. making it text only like lynx

[true ridge]

no html images?

[willow ginkgo]

lol

i mean i could load images

but the minute you do that you then have to go down the css rabbit hole

[true ridge]

just do it idk how hard its to port jpeg or whatever

[true ridge]

oh god

[willow ginkgo]

and i dont have a spare extra life of time to make a graphical web browser

[true ridge]

every time html and css get mentioned i get ptsd from my geniunely braindead web design teacher i swear

[willow ginkgo]

a text only browser where you can tab into an image would work

[true ridge]

why did i even sign up there

[true ridge]

separate app?

just open it like a separate file or smth

[willow ginkgo]

hmmm, image viewer app, thats not a bad idea

easy to do too

[true ridge]

how difficult is png to implement

[willow ginkgo]

just spriteload it

idk, you think i wasted my time reinventing that wheel? lol

[true ridge]

im pretty sure its lossless so its probably just "red pixel 50 times, green pixel 20 times"

[true ridge]

sprites

[willow ginkgo]

i use stb_image

stb_image can load tons of different formats

including png, gif, bmp, etc

[true ridge]

just looked it up

looks like it's worth porting

except im big stupid so idk

[willow ginkgo]

i dont get why so many people here spend so long trying to implement png, unless they do it purely to learn

it doesnt need any porting, it just works

[true ridge]

oh

copy paste it

[willow ginkgo]

single header only lib

yup

[true ridge]

idk theres a ton of folders and stuff in it it just looks menacing as absolute shit

[willow ginkgo]

nah

[true ridge]

fair

[willow ginkgo]

stb_image.h is legit all you need

[true ridge]

fair™

[willow ginkgo]

https://tenor.com/view/fair-that's-fair-stefon-bill-hader-snl-gif-11240744675777193136

[true ridge]

also is it difficult to implement a 3d library because that 3d cube goes hard as shit (also how does it render so fast without graphics drivers)

[willow ginkgo]

the simple stuff is simple

what ive done isnt much beyond high school maths

[true ridge]

ok fair

[willow ginkgo]

but if you want the more advanced stuff it can get a bit hairy

nomal mapping, shadow maps, reflections, ssr, all that

[true ridge]

i think i did a 3d implementation at some point

[willow ginkgo]

anywhere else i'd use an existing engine, but that isnt happening in retro rocket

[true ridge]

Oh Dear God Just Use Windows At That Point™

[true ridge]

understandable

[willow ginkgo]

my preferred poison is unreal engine

ive released games before

[true ridge]

unrelated question but how the hell does roblox have unreal engine level graphics
it's like there's 1 year left until they run ray tracing on a 10 year old ipad (like literally if you try hard enough you'd get rtx on a powermac g3)

i mean i dont really see the difference

there def is but not major

[true ridge]

:0

[willow ginkgo]

i think tomorrow i'll make a hosts file

by that i mean support for it

[true ridge]

a what file

what does that do

[willow ginkgo]

/etc/hosts or c:\windows\system32\drivers\etc\hosts

its a set of configured aliases that override dns

you never used it?

[true ridge]

oh

[true ridge]

no

https://tenor.com/view/dead-my-honest-reaction-damn-skull-skeleton-gif-15885937969029652188

i feel like i just skipped a half of using a personal computer in general

[willow ginkgo]

adding some simple command line tools. i need these for searching the /devices/debug

[heavy linden]

That BASIC seems pretty ADVANCED imo /s

[willow ginkgo]

lol, this is true

except its only as advanced as you make it

IF ARGS$ = "" THEN
    PRINT "File name not specified"
    END
ENDIF
FILE$ = TOKENIZE$(ARGS$, " ")
SEARCH$ = LOWER$(ARGS$) ' Everything that remains
FH = OPENIN(FILE$)
IF FH < 0 THEN
    PRINT "Unable to open file: "; FILE$
    END
ENDIF
LS$ = ""
LN = 1
F = 0
REPEAT
    BYTE = READ(FH)
    LS$ = LS$ + CHR$(BYTE)
    IF BYTE = 10 THEN
        IF INSTR(LOWER$(LS$), SEARCH$) > 0 THEN
        PRINT LN; ":", LS$;
            F = F + 1
        ENDIF
        LS$=""
        LN = LN + 1
    ENDIF
UNTIL EOF(FH) = 1
CLOSE FH
IF F > 0 THEN
    PRINT ""
    PRINT F; " matches found."
ELSE
    PRINT "No matches found."
ENDIF

thats the search program

[heavy linden]

(Im just gonna act like a understand all of this.)

[willow ginkgo]

lol its easy

it does exactly what the keywords look like it does

read bytes from a file until it has a line, and if the line contains the search string print it to the screen

[willow ginkgo]

feels very much like old school basic now

just with immense amounts of ram

[short token]

basic is the worst programming language ever tbh

[willow ginkgo]

nah that would be brainfuck

pretty much if you wanted to learn to program up until the early 90s you learned BASIC

[true ridge]

@grok is this true

[willow ginkgo]

btw the images above, ive changed the way heap is initialised so that even with identity mapped memory, i can claim all the 'parts' and coalesce them into one managed heap i can call from via kmalloc

generally, youll gain a gigabyte if your vm or real machine has anything more than 2gb ram

and always about a gigabyte

because everything below 4gb is punctuated by 'holes' of unusable memory

acpi, mmio, etc

[true ridge]

nice

[willow ginkgo]

better error handler, for when it goes crazy

[latent tendon]

Damn thats cool

How do u know which context it is?

And how do u backtrace through irqs

[willow ginkgo]

i am placing a sentinel value into the stack

its pushed into the trace stack before i call IRQ or Interrupt, from the assembly stub

the interrupt call and the kernel itself share a stub, this makes it easier to cross contexts than if it was a 'user context' in ring 3 with a different stack

when i see the sentinel value i know that we entered interrupt context

i count the sentinels in the stack before i draw the stack trace, and decrement each time i cross one, when i cross the last one, i know thats pre-empted non-interrupt code

static size_t count_markers_ahead(const stack_frame_t *frame, uintptr_t lo, uintptr_t hi, size_t probe) {
    size_t cnt = 0;
    while (frame && CANONICAL_ADDRESS(frame) && (uintptr_t)frame >= lo && (uintptr_t)frame <= (hi - sizeof(*frame)) && probe++ < 1024) {
        const stack_frame_t *next = frame->next;
        if (next && (uintptr_t)next <= (uintptr_t)frame) {
            break;
        }
        if (frame->addr == BT_IRQ_MARKER) {
            cnt++;
        }
        frame = next;
    }
    return cnt;
}

void backtrace(void) {
    stack_frame_t *frame;
    __asm__ volatile("movq %%rbp,%0" : "=r"(frame));
    uintptr_t lo, hi;
    size_t depth = 0;
    get_kstack_bounds(&lo, &hi);

    /* We know we’re in an ISR/exception at the top */
    setforeground(COLOUR_LIGHTYELLOW);
    aprintf("--------------------------------[ IRQ/TRAP CONTEXT ]--------------------------------\n");
    size_t remaining_markers = count_markers_ahead(frame, lo, hi, depth);
    setforeground(COLOUR_LIGHTGREEN);
    while (frame && CANONICAL_ADDRESS(frame) && (uintptr_t)frame >= lo && (uintptr_t)frame <= (hi - sizeof(*frame)) && depth++ < 1024) {

        stack_frame_t *next = frame->next;

        if (next && (uintptr_t)next <= (uintptr_t)frame) {
            break;
        } else if (frame->addr == BT_IRQ_MARKER) {
            if (remaining_markers > 0) {
                remaining_markers--;
            }
            setforeground(COLOUR_LIGHTYELLOW);
            if (remaining_markers > 0) {
                aprintf("--------------------------------[ IRQ/TRAP CONTEXT ]--------------------------------\n");
            } else {
                aprintf("-------------------------------[ PRE-EMPTED CONTEXT ]-------------------------------\n");
            }
            setforeground(COLOUR_LIGHTGREEN);
            frame = next;
            continue;
        }
        uint64_t offset = 0;
        const char *name = findsymbol((uint64_t)frame->addr, &offset);
        aprintf("\tat %s()+0%08lx [0x%lx]\n", name ? name : "[???]", offset, (uint64_t)frame->addr);
        frame = next;
    }
    setforeground(COLOUR_WHITE);
}

for the debug log, i save dprintf() to a circular buffer, i use a call to take a snapshot of the buffer and then output the last n lines of it

for the asm:

.macro PUSH_SENTINEL
    sub $16, %rsp
    mov %rbp, 0(%rsp)
    movq $BT_IRQ_MARKER, 8(%rsp)
    mov %rsp, %rbp
.endm

.macro POP_SENTINEL
    mov 0(%rbp), %rbp
    add $16, %rsp
.endm

interrupt_stub:
    PUSH_SENTINEL
    call Interrupt
    POP_SENTINEL
    MPOPA
    iretq

[latent tendon]

Ohh

Cool

[willow ginkgo]

now made it able to find symbols in loaded modules, too 😄

[true ridge]

make a basic instruction for kernel panic

[latent tendon]

wait so how does the code above handle irq frames vs normal frames?

since rsp/rip are at different offets no?

[willow ginkgo]

there is/was one

it I disabled it I think

[willow ginkgo]

nope same stack at least for me

probably because it doesnt switch privilege contexts

[latent tendon]

ah wait

[willow ginkgo]

I don't switch between ring 0 and 3, any permission checks are enforced by the interpreter... that which there are. not many rn

[latent tendon]

wouldnt this mean you're discarding the actual interrupted function?

since your moved rbp is of the previous caller

[willow ginkgo]

the idea is that a basic program can remove privileges on a granular basis from children "next child can't use keyword x, y z" like a chroot

[willow ginkgo]

doesn't seem so?

this is the first time I've revisited my stack trace code in about a decade if not more, so I might be missing something too

I just use the standard frame pointers GCC emits

and it seems to follow execution into interrupts

[latent tendon]

so like

rip of caller of caller
rbp of caller of caller
----
rip of caller
rbp of caller
---
rbp of caller
your marker

i dont see whre the rip of the called function shows up here

maybe im missing something

[willow ginkgo]

I don't replace anything I push a new frame entirely

[latent tendon]

yes but rbp still contains the old base value right

[willow ginkgo]

the new frame has nothing but a sentinel rip, 0xffffffffffffff01

if it's found it's skipped and the header printed

[latent tendon]

so how does it know the rip of the interrupted function

[willow ginkgo]

the frame contains next, sentinel

[latent tendon]

since next should be the base of the interrupted function

which has the rip of the caller

[willow ginkgo]

that's still there, the sentinel is pushed inside the stub

[latent tendon]

the sentinel is not a rip though right

its just a thing to print the header

[willow ginkgo]

no it is not, not a valid one

yup

[latent tendon]

so yeah i still dont get it

or hmm maybe the interrupted rip is next to your pushed rbp

so like

... other values pushed by CPU
interrupted cs
interrupted rip
your stub rbp
your marker

i guess is the stack layout so it works?

[willow ginkgo]

looks right, I'm struggling to visualise what you mean

[latent tendon]

anyway if it works it works lol

[willow ginkgo]

I insert an entirely spoofed stack frsme, right before calling into the irq handler (which makes a new frame itself), then immediately after pop the spoofed one

the irqs are dispatched by 256 macros, basically a bit like the old jamesm tutorial just fixed and adapted for 64 bit and extended to support the full range of interrupts for msi

so initially the interrupt goes to that, which is a macro, which pops the sentinel

[latent tendon]

my question is

if the frame is spoofed

how is the next frame correct

[willow ginkgo]

then it sets the irq number and interrupt number parameter and calls the global function IRQ or Interrupt

I don't do anything to the next frame

[latent tendon]

in my understanding with this approach you discard the interrupted function and actually start the backtrace from its caller

unless you pull the rip out of the stack frame somewhere separately

[willow ginkgo]

so it's like irq0 -> interrupt_stub -> push sentinel -> call IRQ() -> pop sentinel -> iret

[latent tendon]

i understand that bit yeah

[willow ginkgo]

oh I see... nope the stack trace goes all the way back to kmain_bsp(), the first function on the stack

[latent tendon]

nah thats not what i meant

[willow ginkgo]

I don't see what's lost or discarded here? I think we are both thinking of different things

[latent tendon]

is the stuff after the second yellow line here correct? Most likely its actually 5 frames deep but you're only showing 4

[willow ginkgo]

what, from dns_lookup_host?

[latent tendon]

handle_receive

[willow ginkgo]

yes that's right

[latent tendon]

is it a leaf function?

[willow ginkgo]

I put a memcpy to null in that function that fires if the system uptime is greater than 5 secs then ran a http test

[latent tendon]

hmm

[willow ginkgo]

it's the immediate function called by e1000_handler in this case attached to irq11

so nothing missing there

[latent tendon]

yeah i dont understand then

[willow ginkgo]

the only oddity is idt_init(), neither interrupt stacks pass through there... but it's the "nearest" symbol to the assembly stubs that don't have symfile entries

technically that is like irq11_stub or something

[latent tendon]

ah wait

u said u put a memcpy

[willow ginkgo]

I could fix that

yup

[latent tendon]

why dont we see the memcpy in the trace

[willow ginkgo]

but my memcpy is inlined

[latent tendon]

can u just *(int*)0x123 = 123?

[willow ginkgo]

should that show?

[latent tendon]

in that function

[willow ginkgo]

yeah I could but the memcpy stopped my linter bitching lol

[latent tendon]

and check what will be printed

[willow ginkgo]

I can't rn, but I'll check later

the only one discarded should be the very first frame (it's all one loop) because that is always just backtrace() itself

pointless to show that lol

[latent tendon]

yeah thats whatever

[willow ginkgo]

to be honest could discard the entire top section and error traces would still make sense

[willow ginkgo]

youre right, the frame alongside the sentinel is omitted, but i think it might always have been omitted. i'm looking into it

[willow ginkgo]

i see what you mean

the fix is simple enough

the fault address doesnt show in the trace, because the trace traces return addresses

theres no return address from where the fault occurs

im working around it by pushing rip onto the stack, and retrieving it and loading it into a parameter that i pass to the interrupt function

i can then use that rip to print the faulting function name, before the trace

[latent tendon]

Yup that's what I meant

[willow ginkgo]

there

also added more detail about the error code if any

[true ridge]

even more detail

reminds me of windows nt 4.0

or other early nt's

[willow ginkgo]

a checked build maybe

[true ridge]

maybe

[latent tendon]

I dont think it ever printed backtraces

It worked via dumps

[willow ginkgo]

now excluding the error handlers context

[willow ginkgo]

moving back now to improving the UX of the editor

[willow ginkgo]

syntax highlighting!

[true ridge]

finally

[willow ginkgo]

its very funky

[true ridge]

colorful

[willow ginkgo]

this actually crosses off a roadmap item

the editor is now usable

[true ridge]

only when you disable the feature it is

everyone knows suffering is key to success

[willow ginkgo]

lol

anger leads to hate

hate leads to ....suffering

https://tenor.com/view/fear-is-the-path-to-the-dark-side-fear-leads-to-anger-anger-leads-to-hate-hate-leads-to-suffering-fear-leads-to-anger-anger-leads-to-hate-hate-leads-to-suffering-gif-13858055100705051300

[willow ginkgo]

Next roadmap item:

Create a simple network file copy (FTP client? something else?) to get files on and off the installed system into github

[true ridge]

OWL

ETSFEIEJDJSJEJELQ39

[true ridge]

good luck

[willow ginkgo]

made a better image with CUUUUUUUUUUUUUBE

[true ridge]

CUBEEEEEEEEEEE

[willow ginkgo]

heap init now reclaims more. we reclaim nearly all bootloader reclaimable.

[true ridge]

isnt that just a few mb

[heavy linden]

thtas about 4gb

[willow ginkgo]

i got a better figure... one sec

[willow ginkgo]

is ....32gb enough for BASIC

thats my test machine where i boot it off the hard disk

[heavy linden]

that should be enough for many applications.

how much ram can RROS handle?

[willow ginkgo]

i didn't set a limit

so as much as your machine can handle

[fossil pagoda]

Bro, your project is amazing! I love it!

[short token]

why

just buy a crappy 2010 laptop if you want a test machine

[wild sedge]

why would he buy another test machine if he already has a test machine

that makes no sense

[short token]

turn the test machine into a non test machine

[wild sedge]

why?

[short token]

you're a glowie if you don't agree with me

[wild sedge]

uh no?

maybe it's his old workstation, i test bentobox on my laptop that has 24gb of ram

[short token]

I can see your pfp glowing through my monitor

[wild sedge]

??

[true ridge]

no the reclaimable areas

[willow ginkgo]

its too old and slow for that

AMD FX-6300, Gigabyte-78LMT-USB3, 32gb ram, 1tb HDD

it was my old pc, yes

[true ridge]

32 gb of ram
"slow"

[willow ginkgo]

its only got 6 cores

[true ridge]

oh

ok thats kinda ass for today

[humble bison]

dual channel ddr3 really isnt the fastest

[willow ginkgo]

yup thats what its got

the board is maxed out

[humble bison]

the issue is more the ram speed than the amount i believe

[willow ginkgo]

i run a lot of memory hungry things at the same time, my current pc has 32gb ram too, and my server has 64

and yeah the bandwidth is too slow for gaming

[heavy linden]

I main a hexacore cpu 💀

[willow ginkgo]

at the time it was stonks

[true ridge]

yeah

idk why i said its ass i reconsider