binji / binjgb Goto Github PK

Hi! I want to use this to embed an emulator on my HTML page and have it pre-load a rom of my choice.

Could you guide me as to how to do this? Thanks!

The scaling is uneven, possibly because of the 1-pixel black border around the screen.

Steps to reproduce

1. Download gb240p.gb from 240p-test-mini v0.20
2. In binjgb, load and run gb240p.gb
3. On the second page of the menu, open Full screen stripes
4. Set magnification to 2x or 3x
5. Press X (bound to the A Button) several times to rotate the stripes

Observed behavior:

One stripe about 1/4th of the way from each side is one pixel narrower than the others.

Expected behavior:

All stripes have the same thickness.

Hi there!

Let me begin by again THANKING you for helping me get this up and running on my site!

I did just run into one issue though. On Chromebooks the save file sometimes fails to load, which forces one to restart their game from the start.

Do you think you know why this is happening, and could you possibly add a patch that would allow one to save their save state as a .sav file that could be played on other emulators (or even sent to other computers)?

Thanks and let me know :D

According to TCAGBD section 5.1.1.

In the latest revision in function emulator_delete, file_data_delete is called to release ROM data.

The data is created outside the emulator, but it's released inside it, which causes unmatched malloc/free. There should be a way that allows a programmer to indicate whether the memory of ROM data is managed by emulator or its caller.

GBG knows two kinds of debug messages that can be used to trace what the program is doing. They're described in the manual.

In the first kind, the message is embedded in the instructions

    ld d, d ; signals a debug message
    jr .enddebugmsg
	; the next 4 instructions specify the kind of mesage
    ld h, h
    ld h, h
   	dw $0000
    db "my message"
.enddebugmsg

The second kind loads a zero-terminated string from an address, which can be useful for dynamic messages that include runtime values.

    ld d, d
    jr .enddebugmsg
    ; the next 4 instructions specify the kind of mesage
    ld h, h
    ld h, h
    dw $0001
    dw address
    dw bank
.enddebugmsg

In BGB, these messages are displayed in a separate window.

This would be a nice feature, I think.

I was seeing an audio latency issue on a page with the emulator embedded, in cases when the audio context starts out in "suspended" state.

As you know, Web Audio API has an autoplay policy which can require a user gesture, such as click or keypress event, and calling resume() on the audio context, before it's in a "running" state.

The issue was that, because I'm starting the emulator immediately upon page load, audio buffers were being created while the audio context was still suspended. This happens in the Audio class, the method pushBuffer(). Every time Audio.ctx.createBuffer is called, it produces a console warning:

An AudioContext was prevented from starting automatically. It must be created or resumed after a user gesture on the page.

Then, when the audio context is "resumed" after a user gesture, it apparently starts processing all the buffered audio, causing significant audio latency in the emulator.

The solution was to avoid creating an audio buffer when the audio context is suspended.

Before this condition:

if (this.startSec >= nowSec) {
  const buffer = Audio.ctx.createBuffer(2, AUDIO_FRAMES, this.sampleRate)
  ...
}

I added a line:

if (Audio.ctx.state !== 'running') return

This is probably too small for a pull request, but I thought it could be a possible improvement for the emulator.

Just wanted to add, I'm thoroughly enjoying the emulator and its code base. Such wonderful work!

[anders/fix_compile_errors_on_osx] % ./bin/binjgb ~/Projects/blarggs-tests/cpu_instrs/source/test.gb
title: ""
cgb flag: CGB_FLAG_SUPPORTED
sgb flag: SGB_FLAG_NONE
cart type: CART_TYPE_MBC1
rom size: ROM_SIZE_32K
ext ram size: EXT_RAM_SIZE_NONE
header checksum: 0x66 [OK]
GL_VERTEX_SHADER ERROR: ERROR: 0:1: '' :  version '130' is not supported
ERROR: 0:2: '' :  #version required and missing.
ERROR: 0:2: 'attribute' : syntax error: syntax error

zsh: segmentation fault  ./bin/binjgb ~/Projects/blarggs-tests/cpu_instrs/source/test.gb

It appears to originate in the hard coding of the GL_VERTEX_SHADER version: https://github.com/binji/binjgb/blob/master/src/host-ui-imgui.cc#L116

However bumping that to version 150 which I believe is the current version available on my machine moves the error to an attribute syntax error:

GL_VERTEX_SHADER ERROR: ERROR: 0:2: 'attribute' : syntax error: syntax error

The Lawnmower Man for Game Boy reportedly will not start unless joypad interrupts are implemented.

To reproduce:

1. Download and unzip the attached test ROM "Telling LYs?" tellinglys-for-binjgb.zip
2. In binjgb, load and run the ROM
3. Press any button

Observed behavior:

Title screen does not disappear; the test cannot progress.

Expected behavior:

Display advances from title screen to conversation, and pressing other buttons advances the conversation.

(If you do issue joypad interrupts, but you don't vary the LY when they're issued, you'll get a different error message.)

Is there a way to run the DMG Boot Rom? I am developing an emulator on a Mac and the debugger looks great but I can't find a way to run the boot rom to compare my own emulator with.

In CGB mode, sprite priority is determined by OAM address (sprite #00 being topmost, #3F being bottommost). binjgb currently uses the DMG behavior of priority being assigned by X position.

• model new registers
• speed switch
• banked vram
• banked wram
• hdma
• new palettes (bg/obj)
• new PPU renderer
• try to keep it under 5k lines still!

Other stuff too, I'm sure. But this is a good start.

Hi Ben,

Could you make some change in your emulator to you can enable/disable logging by pressing a key? I mean then I run emulator with '-t' flag it starts to write in trace log, but I need to get trace log in specific time in the game. It would be nice to enable or disable logging to file by pressing a key.

I cloned the repository and run make. Make failed because imgui was not present, i cd into third_party/ and cloned the imgui (https://github.com/ocornut/imgui.git). Then i run make again. This time it started to compile but it exited with these errors:

/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc: In member function ‘bool Debugger::Init(const char*, int, int, int, bool, u32, u32, bool)’:
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:127:21: error: ‘ImGuiConfigFlags_DockingEnable’ was not declared in this scope; did you mean ‘ImGuiConfigFlags_None’?
  127 |   io.ConfigFlags |= ImGuiConfigFlags_DockingEnable;
      |                     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
      |                     ImGuiConfigFlags_None
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:128:6: error: ‘struct ImGuiIO’ has no member named ‘ConfigDockingWithShift’
  128 |   io.ConfigDockingWithShift = false;
      |      ^~~~~~~~~~~~~~~~~~~~~~
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc: In member function ‘void Debugger::Run()’:
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:241:18: error: ‘DockBuilderGetNode’ is not a member of ‘ImGui’
  241 |       if (ImGui::DockBuilderGetNode(dockspace_id) == nullptr) {
      |                  ^~~~~~~~~~~~~~~~~~
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:242:16: error: ‘DockBuilderRemoveNode’ is not a member of ‘ImGui’
  242 |         ImGui::DockBuilderRemoveNode(dockspace_id);
      |                ^~~~~~~~~~~~~~~~~~~~~
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:243:16: error: ‘DockBuilderAddNode’ is not a member of ‘ImGui’
  243 |         ImGui::DockBuilderAddNode(dockspace_id, ImGui::GetIO().DisplaySize);
      |                ^~~~~~~~~~~~~~~~~~
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:249:16: error: ‘DockBuilderSplitNode’ is not a member of ‘ImGui’
  249 |         ImGui::DockBuilderSplitNode(dockspace_id, ImGuiDir_Left, 0.333f, &left,
      |                ^~~~~~~~~~~~~~~~~~~~
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:251:16: error: ‘DockBuilderSplitNode’ is not a member of ‘ImGui’
  251 |         ImGui::DockBuilderSplitNode(left, ImGuiDir_Up, 0.666f, &left_top,
      |                ^~~~~~~~~~~~~~~~~~~~
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:253:22: error: ‘DockBuilderSplitNode’ is not a member of ‘ImGui’
  253 |         mid = ImGui::DockBuilderSplitNode(mid, ImGuiDir_Left, 0.5f, nullptr,
      |                      ^~~~~~~~~~~~~~~~~~~~
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:255:16: error: ‘DockBuilderSplitNode’ is not a member of ‘ImGui’
  255 |         ImGui::DockBuilderSplitNode(mid, ImGuiDir_Up, 0.5f, &mid_top,
      |                ^~~~~~~~~~~~~~~~~~~~
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:258:16: error: ‘DockBuilderDockWindow’ is not a member of ‘ImGui’
  258 |         ImGui::DockBuilderDockWindow(s_emulator_window_name, left_top);
      |                ^~~~~~~~~~~~~~~~~~~~~
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:259:16: error: ‘DockBuilderDockWindow’ is not a member of ‘ImGui’
  259 |         ImGui::DockBuilderDockWindow(s_audio_window_name, left_bottom);
      |                ^~~~~~~~~~~~~~~~~~~~~
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:260:16: error: ‘DockBuilderDockWindow’ is not a member of ‘ImGui’
  260 |         ImGui::DockBuilderDockWindow(s_rewind_window_name, left_bottom);
      |                ^~~~~~~~~~~~~~~~~~~~~
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:261:16: error: ‘DockBuilderDockWindow’ is not a member of ‘ImGui’
  261 |         ImGui::DockBuilderDockWindow(s_obj_window_name, mid_top);
      |                ^~~~~~~~~~~~~~~~~~~~~
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:262:16: error: ‘DockBuilderDockWindow’ is not a member of ‘ImGui’
  262 |         ImGui::DockBuilderDockWindow(s_tiledata_window_name, mid_top);
      |                ^~~~~~~~~~~~~~~~~~~~~
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:263:16: error: ‘DockBuilderDockWindow’ is not a member of ‘ImGui’
  263 |         ImGui::DockBuilderDockWindow(s_map_window_name, mid_bottom);
      |                ^~~~~~~~~~~~~~~~~~~~~
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:264:16: error: ‘DockBuilderDockWindow’ is not a member of ‘ImGui’
  264 |         ImGui::DockBuilderDockWindow(s_disassembly_window_name, right);
      |                ^~~~~~~~~~~~~~~~~~~~~
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:265:16: error: ‘DockBuilderDockWindow’ is not a member of ‘ImGui’
  265 |         ImGui::DockBuilderDockWindow(s_memory_window_name, right);
      |                ^~~~~~~~~~~~~~~~~~~~~
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:266:16: error: ‘DockBuilderDockWindow’ is not a member of ‘ImGui’
  266 |         ImGui::DockBuilderDockWindow(s_io_window_name, right);
      |                ^~~~~~~~~~~~~~~~~~~~~
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:267:16: error: ‘DockBuilderDockWindow’ is not a member of ‘ImGui’
  267 |         ImGui::DockBuilderDockWindow(s_rom_window_name, right);
      |                ^~~~~~~~~~~~~~~~~~~~~
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:268:16: error: ‘DockBuilderFinish’ is not a member of ‘ImGui’
  268 |         ImGui::DockBuilderFinish(dockspace_id);
      |                ^~~~~~~~~~~~~~~~~
/home/ucak/Projects/emulator/binjgb/src/debugger/debugger.cc:271:14: error: ‘DockSpace’ is not a member of ‘ImGui’
  271 |       ImGui::DockSpace(dockspace_id);
      |              ^~~~~~~~~
make[3]: *** [CMakeFiles/binjgb-debugger.dir/build.make:232: CMakeFiles/binjgb-debugger.dir/src/debugger/debugger.cc.o] Error 1
make[2]: *** [CMakeFiles/Makefile2:171: CMakeFiles/binjgb-debugger.dir/all] Error 2
make[1]: *** [Makefile:130: all] Error 2
make: *** [Makefile:15: debug] Error 2

Description

When Fatass Tracker tries to load, it doesn't work.
It says "File size too small" and freezes.
However, with other emulators, that rom perfectly works!

Steps

1. Go here: https://hh.gbdev.io/game/fatass
2. Open web console
3. It says "File size too small (required 524288, got 425984)"
4. (optional) to test yourself, download the rom, download mgba. It works.

Could we do something to fix this behaviour?

Since commit number b18c9ac (Synchronize video to the monitor refresh rate) committed on Mar/20/2017, the Windows build runs much faster than it should.

I have tried your released binaries and I have also compiled the source code myself. Have run them on my machine and also on a virtual machine. Happens all the time.

If you haven't seen it on Linux or Mac, I'm guessing that this only happens when using the Windows build of SDL 2.

I was playing "Super Mario Bros. Deluxe (USA, Europe)" using the emulator in the website and noticed that it's using lots of RAM

I think I spotted a bug. :) Please take a look at A value and flags after add a,a at 0x01f7:

A:80 F:---- BC:0000 DE:c205 HL:0260 SP:fffc PC:01ec (cy: 548) ppu:+3 |[00]0x01ec: 18 09     jr +9          
A:80 F:---- BC:0000 DE:c205 HL:0260 SP:fffc PC:01f7 (cy: 560) ppu:+3 |[00]0x01f7: 87        add a,a        
A:00 F:---C BC:0000 DE:c205 HL:0260 SP:fffc PC:01f8 (cy: 564) ppu:+3 |[00]0x01f8: 38 08     jr c,+8        
A:00 F:---C BC:0000 DE:c205 HL:0260 SP:fffc PC:0202 (cy: 576) ppu:+3 |[00]0x0202: 28 ea     jr z,-22       
A:00 F:---C BC:0000 DE:c205 HL:0260 SP:fffc PC:0204 (cy: 584) ppu:+3 |[00]0x0204: 01 02 00  ld bc,2        

Carry is set but Zero is missing so instead making conditional jump at 0x0202 it goes straight to 0x0204 which is wrong. I checked it in BGB before opening an issue and it sets Z--C so the result is as expected. Above output is from binjgb-tester.exe win64. This is really handy util BTW!

when i click on the link to a copy i made, it appears as a black screen
this is the page, if you click on-site test (number) then it will take you to the embed
https://teddblue.github.io/gamebois/GBC.html
this is the full project with a link to the site
https://github.com/teddblue/gamebois

I think I know how this works now, partly via Nitty gritty VRAM doc, and partly from the sprite tests in mooneye-gb. I don't think I'll have to implement the full pipeline behavior, because it's not observable outside the fetches from VRAM. (Though maybe that's not observable?)

334B is the minimum size needed. 334th Byte is the header checksum, which is the last header byte touched by the bootrom. Global checksum doesn't get verified. SIERPINSKIBOY should therefore have everything, the bootrom needs.

$ bin/binjgb testrom/sierpinskiboy.smol.cgb
/tmp/binjgb/src/emulator.c:4847: File size (334) should be a multiple of minimum rom size (32768).

TINYMIRRORBOY has the correct first half of the Nintendo logo and a correct header checksum. That's everything CGB needs to boot.

$ bin/binjgb testrom/tinymirrorboy.mirrored.32k.cgb
/tmp/binjgb/src/emulator.c:1178: Invalid ROM.

These ROMs work on Sameboy, Emulicious and BGB. And they also boot with EVERDRIVE GB X3 on my CGB-D.

See main...daid:binjgb:master.

This will make it easier to use upstream binjgb for rgbds-live

Description

When Cube Raider tries to load, it loads, but it's not correctly viewed.
This is a behavior encountered with other emus too.

Steps

• Go here: https://hh.gbdev.io/game/Cube-Raider
• What should happens is described in this video
• (optional) to test yourself, download the rom, download mgba. Same behavior.

Could we do something to fix this behaviour?
TBH I don't know if we can do anything for it, but it would be nice to have it working.

They're trickier than I have currently: e.g. if you DMA from VRAM to OAM, it shouldn't block the CPU (read this somewhere, can't remember where...)

Also, it should be possible to restart a DMA, so access to FF46 should be legal during DMA.

The current implementation passes Blargg's tests, but probably not gekkio's.

Calling SDL_Init(SDL_INIT_EVERYTHING) is not needed and will fail on some systems that doesn't have all SDL components installed.

I couldn't find it documented, so I thought I'd ask: Is there any way to force DMG mode? I'm trying to use your emulator's tracing to see where my emulator diverges, but I'm not currently aiming for CGB support.

It doesn't seem like the memory is "random" enough when uninitialized after boot. This means that common PRNG routines that rely on uninitialized RAM as a seed don't work.

Hi,
I have a trouble: when I launch binjgb-debugger nothing happens. There is music from Game, but screen is black (gray).
When I press some keys there are some responds, but screen stays in black.

As per these gbdev.io docs, when it comes to drawing an OAM pixel, the Gameboy first calculates which of 2 overlapping objects has higher priority; and THEN considers the winning object's BG priority bit. This can be exploited to "mask" objects (see INTERACTION WITH "BG OVER OBJ" FLAG in the article.)

Currently on binjgb, this behavior does not occur; it calculates the BG priority bit of all objects and renders whatever is on top.

On my local version of binjgb, I was able to fix this by inserting the following lines to emulator.c:



which seems like a hacky solution, but you can see a working version on my demo GB game.

I'm unable to open a pull request as I'm unable to create a branch but this is the issue and a potential solution!

Don't call SDL_PauseAudio(0) until there is more data available.

Some other underflows happen occasionally too, maybe those are from locking on the audio thread? Need to investigate.

When you fly down the screen, there is a spot where the window fills the entire screen. The issue is that the LY=LYC interrupt should be firing, but isn't. The strange thing is that normally the code sets the LYC value to something below the current LY value. When it screws up, it sets LYC to the current LY value. This should fire an interrupt, but it seems to be blocked by the internal STAT interrupt flag already being set (see stat_irq_blocking in mooneye-gb). Maybe I'm setting it wrong.

The rgbds toolchain (https://github.com/rednex/rgbds) can output a .sym file, where all labels used in the assembler code are given with their location. It looks something like this.

; this is a comment
; the format is
; bank:address label
00:08f4 Start.mainloop
00:09b0 memcpy
00:09b4 memcpy.repeat
00:09b7 memcpy.start

The bgb emulator reads the file if it has the same name as the rom file (except the file extension is .sym) and displays the labels in the disassembly, which looks like this:

The labels are very helpful while debugging and it would be awesome if binjgb could support it.

If a game has the BG and window disabled, but objects enabled, pixels from sprites linger once drawn until overwritten by sprites once again.

I tried by different means to download the 'db' indexedDB data.
I was able to see the base64-encoded screenshot, but the 'extram' blob, despite having a size of 131072, was always empty.

Is there anything special to do in order to import or export .sav files?

thx