00:00:00 --- log: started retro/14.12.06
07:48:17 --- join: mnemnion (~mnemnion@c-98-210-219-91.hsd1.ca.comcast.net) joined #retro
10:27:05 --- quit: Sgeo_ (Read error: Connection reset by peer)
11:07:21 --- join: Sgeo (~quassel@ool-44c2aebc.dyn.optonline.net) joined #retro
12:35:04 --- quit: mnemnion (Remote host closed the connection)
15:16:16 --- join: FreeFull (~freefull@defocus/sausage-lover) joined #retro
15:17:09 <FreeFull> I've been developing an implementation of the Ngaro Virtual Machine based on http://retroforth.org/docs/The_Ngaro_Virtual_Machine.html
15:17:13 <FreeFull> It's not ready to use yet
15:18:00 <tangentstorm> hey! welcome, FreeFull :)
15:18:02 <tangentstorm> rust?
15:18:08 <FreeFull> Heya tangentstorm
15:18:09 <FreeFull> Yes
15:18:36 <FreeFull> I'll need to test if I implemented the CPU correctly, but I need to implement IO first to do that
15:19:18 <FreeFull> What is the expected behaviour when a stack is empty, but an instruction calls for popping a value off it?
15:19:27 <tangentstorm> if you implement saving the vm back to disk, there's a test script written in python that you can use
15:19:35 --- join: mnemnion (~mnemnion@c-98-210-219-91.hsd1.ca.comcast.net) joined #retro
15:19:39 <FreeFull> Thanks
15:19:40 <tangentstorm> ngaro crashes
15:19:54 <FreeFull> Ok, I probably should do that too
15:19:56 <tangentstorm> (and then once you have those working, there's another test suite written in retro that runs on the vm)
15:20:31 --- join: SkyTheLeafeon (ae3d85bd@gateway/web/freenode/ip.174.61.133.189) joined #retro
15:20:33 <FreeFull> Right now the implementation I've got just keeps going, which probably isn't a good thing
15:20:37 <FreeFull> I should make it at least stop
15:20:58 <tangentstorm> test/ngaro/ngarotest.py
15:21:00 --- part: SkyTheLeafeon left #retro
15:21:18 <FreeFull> I'll try to get keyboard and text working first
15:21:21 <tangentstorm> it's pretty much undefined behavior. the c implementation just happens to crash.
15:22:19 <tangentstorm> hey docl's here too. long time no see.
15:22:53 <FreeFull> I haven't looked at any other implementation, that document is the only thing I've been using =P
15:23:22 <tangentstorm> fair enough :)
15:25:22 <FreeFull> Why is the text output described as a "Character Generator"?
15:26:43 <tangentstorm> i think a character generator is a hardware device that renders characters graphically. (like the text mode on an ibm pc) not sure.
15:33:37 <FreeFull> Ok
15:42:49 <FreeFull> Doesn't say if it should be ASCII or such
15:43:10 <tangentstorm> nope. characters are 32-bit values
15:44:47 <FreeFull> I mean, for display
15:45:04 <FreeFull> Unless I should only display numbers =P
15:47:18 <tangentstorm> the encoding you use just depends on how you implement the terminal
15:47:52 <tangentstorm> if it's graphical, then you can just have a sparse array of bitmaps for all 2^32 possible character values.
15:48:19 <tangentstorm> or if you're outputting to a conventional terminal, you could utf-8 encode it on the fly or whatveer.
15:48:23 <tangentstorm> lemme see what i did.
15:50:40 <tangentstorm> hrm. i just assumed everything was a byte and utf-8 just magically worked. :)
15:50:47 <FreeFull> I mean, is there a predefined bitmat for each character?
15:50:51 <tangentstorm> nope
15:51:01 <FreeFull> Ok
15:51:13 <FreeFull> I'll just output the u8 value to stdout for now
15:57:12 <FreeFull> Are things like keyboard arrows meant to work?
15:57:29 <tangentstorm> undefined.
15:57:58 <FreeFull> Too bad
15:58:01 <FreeFull> They are really  useful =P
15:58:44 <tangentstorm> you can configure your keyboard device to send whatever codes you want.
15:59:16 <FreeFull> Yes, but that's useless if the software doesn't know what they mean =P
15:59:56 <tangentstorm> yeah but it doesn't not know what they mean either.
16:00:15 <tangentstorm> in retro, you can say getc and press any key and it'll read that character.
16:00:47 <tangentstorm> there's no line buffering so every keypress is handled directly.
16:01:15 <FreeFull> Yeah, I know that
16:01:18 <tangentstorm> so if you tell your keyboard implementation to send it as one character, you can do that.
16:01:35 <FreeFull> But I'd have to specifically modify/write the software to take advantage of the special codes
16:01:59 <tangentstorm> yep.
16:02:07 <tangentstorm> welcome to retro. :)
16:03:08 <tangentstorm> actually there is already a hook in retro for single-character keyboard macros. you could probably chain those together without too much trouble.
16:03:10 <FreeFull> Why not just add a "standard" set of arrow codes to the standard?
16:03:29 <tangentstorm> i dunno. i guess nobody's needed arrow codes before.
16:03:31 <FreeFull> Say, 257 is left, etc
16:03:51 <tangentstorm> i think the text editor uses vim keybindings.
16:06:24 <tangentstorm> i'm not sure, but i suspect my pascal implementation will send a two-character sequence for arrow keys.
16:08:52 <FreeFull> Mine will end up sending whatever escape sequence the terminal uses
16:09:07 <tangentstorm> that'll work
16:13:44 * crc probably should standardize special (e.g., non-alphanumeric) keys someday
16:14:14 <tangentstorm> heya crc 
16:14:22 <crc> hi tangentstorm 
16:14:53 <FreeFull> Hi crc
16:15:03 <crc> hi FreeFull 
16:15:09 <FreeFull> I'm getting close to having basic IO working
16:15:23 <crc> what language are you writing your implementation in?
16:15:24 <FreeFull> Then I can try to test my code
16:15:26 <FreeFull> Rust
16:15:36 <crc> cool :)
16:43:36 <FreeFull> Ok, I think I have all the bare essentials in place now
16:43:40 <FreeFull> Now to actually get it to work =P
16:45:21 <FreeFull> I've had it print out the IP, and it seems to be looping at the very least
17:00:45 <FreeFull> I've made some changes, and now the instruction pointer is stuck with a value of 7
17:01:46 <tangentstorm> :)
17:02:05 <crc> is the VM_JUMP opcode working? you shouldn’t ever hit 7 (that’s in the start of the default text input buffer IIRC)
17:02:54 <FreeFull> Oops, I accidentally got rid of the line that incremented it
17:02:57 <FreeFull> So it got stuck on a NOP
17:04:17 <FreeFull> Aaand now my vm crashes with a data stack underflow
17:06:10 <FreeFull> ... during a POP instruction...
17:06:43 <FreeFull> Wait, no STORE
17:06:45 <FreeFull> That makes more sense
17:08:40 <FreeFull> Debugging this will be hard ):
17:15:22 <FreeFull> I think I figured it out. My JUMP instruction does the wrong thing
17:17:07 <crc> I have found that once jumps, calls, and basic i/o are working, the rest becomes functional quickly
17:19:39 <FreeFull> Does retroImage require a canvas device?
17:21:12 <crc> no
17:21:15 <crc> canvas is entirely optional
17:21:31 <FreeFull> Ok
17:21:53 <FreeFull> It seems to be stuck in a loop involving port 5
17:22:03 <crc> the only things required to get it working are the keyboard, terminal output, and queries
17:22:43 <crc> port 5 is the capabilities query
17:22:52 <FreeFull> Yes
17:23:13 <crc> it queries:
17:23:19 <crc> i: run-on-boot  (  -  )
17:23:19 <crc>    -1  # query memory # !,  ( Memory Size     )
17:23:19 <crc>    -2  # query fb #     !,  ( Canvas Present? )
17:23:19 <crc>    -3  # query fw #     !,  ( Canvas Width    )
17:23:19 <crc>    -4  # query fh #     !,  ( Canvas Height   )
17:23:20 <crc>    -11 # query cw #     !,  ( Console Width   )
17:23:20 <crc>    -12 # query ch #     !,  ( Console Height  )
17:23:20 <crc>    boot ;
17:23:29 <FreeFull> Yep, that's it
17:23:33 <crc> so, -1, -2, -3, -4, -11, and -12 are needed
17:23:51 <FreeFull> It's stuck in a loop just querying these over and over
17:24:15 <FreeFull> I'm assuming 0 for false, 1 for true?
17:25:19 <FreeFull> I probably should implement -11 and -12 thhen
17:25:19 <crc> -1 for true
17:25:23 <FreeFull> Ah
17:25:36 <crc> does your vm_in instruction clear the port after returning the value?
17:25:50 <FreeFull> Yes
17:26:04 <crc> ok
17:26:16 <FreeFull> Actually, wait
17:26:26 <FreeFull> It doesn't! I must have missed it
17:27:33 <FreeFull> Thanks
17:27:55 <crc> no problem
17:28:47 <FreeFull> Now it seems to get stuck in something that doesn't involve ports
17:32:10 <FreeFull> Or rather, looping and jumping around
17:34:13 <FreeFull> I'm not sure how to debug this
17:34:54 <crc> debugging is painful :(
17:35:31 <crc> I did a lot of trial and error, and lots of printf() (or equiv) in the implementations I wrote
17:36:01 <FreeFull> I think I'll write a function that translates opcodes into instruction names
17:36:05 * crc has intended to put together smaller images for testing specific instructions, but hasn’t done this yet :(
17:36:53 <FreeFull> That'd be nice to have
17:46:20 <FreeFull> https://dpaste.de/nKGs/raw This isn't very meaningful, is it?
17:46:32 <FreeFull> As a debug output
17:47:49 <crc> it’s not much different than what I had while working on a lot of the implementations 
17:48:21 <crc> I think I also displayed the stack depth and top few items 
17:48:34 <FreeFull> I'll do that
17:50:17 <FreeFull> Address Depth seems to keep growing
17:50:40 <FreeFull> How high does it usually get?
17:51:26 <FreeFull> Data Stack depth grows too, but slower
17:52:01 <crc> for the latest image, it gets to 16 to reach the interpreter level, higher when running code
17:52:16 <crc> data stack should be zero once it reaches the interpreter
17:52:48 <FreeFull> IP: 1753 Instr: Some(DEC) Data Depth: 704 Address Depth: 2574
17:52:53 <crc> (the c implementation reports a max depth of 8 for data, 16 for address, when exectuting ‘bye’ immediately after reaching the interpreter'
17:53:09 <crc> so something there isn’t working. 
17:53:17 <FreeFull> Definitely
17:53:20 <crc> is the code online anywhere?
17:53:42 <FreeFull> I'll put it up
17:53:54 <FreeFull> Do you want just the code for the cpu part, or all of it?
17:54:18 <crc> just the cpu (I don’t have rust installed anywhere, but I should be able to look for bugs in that)
17:54:36 <FreeFull> Ok
17:56:12 <FreeFull> https://gist.github.com/FreeFull/0d22991aa75a38f09bf4 Instruction implementations start at line 147
17:57:25 <FreeFull> Also, thank you for helping me out just like that
18:10:30 <crc> I’m happy to help where I can
18:10:48 <crc> nothing jumps out after a quick read of the code :(
18:11:33 <FreeFull> Oh well ):
18:29:52 <FreeFull> You know what
18:30:01 <FreeFull> The implicit call never actually does the jump
18:30:16 <FreeFull> Oh, wait, nevermind, it does
18:31:48 <FreeFull> Oh, oh, I think I know
18:32:15 <FreeFull> The way cond_stack_jump is written, the address ends up being executed as an instruction
18:32:25 <FreeFull> If the condition isn't fulfilled
18:34:16 <FreeFull> Now it waits for keyboard input! \o/
18:34:24 <crc> :)
18:34:48 <FreeFull> And the output seems to work
18:35:02 <FreeFull> Unfortunately, it crashes as soon as I give it any input
18:35:07 <FreeFull> Still, progress
18:38:36 <FreeFull> Seems there is a STORE instruction that tries to store outside of the memory bounds
18:51:59 <FreeFull> It seems to be to some negative address
20:21:53 --- quit: beretta (Quit: Leaving)
23:29:08 --- quit: mnemnion (Remote host closed the connection)
23:44:23 --- quit: crc (Ping timeout: 272 seconds)
23:59:59 --- log: ended retro/14.12.06