NES emulator written in Rust

This kind of got out of hand, but SprocketNES now compiles with the newest Rust nightly, beta and stable.

I've also updated the directory structure (following the Cargo guide), and split the project into a library and a binary (which I'm not sure makes much sense now, but if someone crazy wants to use this in something like a multi-emulator, this will make it like 1% easier).

Oh, and I've enabled V-Sync, which made a massive difference for me (games were massively stuttering before).

I wanted to steal some code for my gameboy emulator but I realized that this code hadn't been updated in a while so I decided to give it a try.

Most of the breakage was due to changes in rust syntax (nampespaced enums, no automatic Copy etc...) however the audio interface to SDL2 seems to have changed quite a lot which means that the current code is completely broken.

In particular rust won't let me declare a static AudioDevice anymore because it implements Drop. I'm not sure how to fix that particular issue without rewriting a big chunk of code so I left it commented for now. If you have any clue what to do I can look into it some more.

But besides that it builds correctly with only one warning left that I can't quite figure out (it looks bogus to me but I might be missing something).

When I tried to run it with mario bros the game seemed very slow even though it wouldn't take all of my CPU, maybe there's a timing issue, I'm not sure (and since I never ran this emulator before I'm not sure how it's supposed to look like).

So, it's not finished but hopefully it'll be useful.

This PR does a few things:

• Update dependencies
• Resolve all warnings
• Fix integer overflows in CPU
• Run rustfmt over the entire codebase

This makes it compile and run again (at least Super Mario Bros.).

Tried twice with some random ROMS:

target/release/nes   MunchieAttack.nes 
Loaded ROM: PRG-ROM: 16 KB, CHR-ROM: 0 KB, Mapper: 0 (0), Trainer: false
thread '<main>' panicked at 'index out of bounds: the len is 0 but the index is 1', src/libcollections/vec.rs:1106
stack backtrace:
   1: 0xb6f837bb - sys::backtrace::tracing::imp::write::h86e39446169e67d0Wnt
   2: 0xb6f860cb - panicking::log_panic::_<closure>::closure.39676
   3: 0xb6f85aef - panicking::log_panic::h7e0ce1e7485944fbUpx
   4: 0xb6f7ae5f - sys_common::unwind::begin_unwind_inner::h1d39a3f983dfda740gs
   5: 0xb6f7b013 - sys_common::unwind::begin_unwind_fmt::h2f77b9962e7023ac6fs
   6: 0xb6f82d97 - rust_begin_unwind
   7: 0xb6f8c79f - panicking::panic_fmt::h48301c5acd531c5bZEK
   8: 0xb6f8c75b - panicking::panic_bounds_check::h8d31f567a438d4e15DK
   9: 0xb6f6bfc7 - mapper::_<impl>::chr_loadb::hb17d08e4fddb2e93FXd
  10: 0xb6f6d5e3 - ppu::_<impl>::step::h122e120144b10602frf
  11: 0xb6f6eb23 - start_emulator::ha09827b57096d9acNGf
  12: 0xb6f66437 - main::h4c1a5b6fa5311b62Qda

and

target/release/nes   SolarWars2001.NES
Loaded ROM: PRG-ROM: 32 KB, CHR-ROM: 32 KB, Mapper: 3 (3), Trainer: false
thread '<main>' panicked at 'unsupported mapper', src/mapper.rs:35
stack backtrace:
   1: 0xb6f9b7bb - sys::backtrace::tracing::imp::write::h86e39446169e67d0Wnt
   2: 0xb6f9e0cb - panicking::log_panic::_<closure>::closure.39676
   3: 0xb6f9daef - panicking::log_panic::h7e0ce1e7485944fbUpx
   4: 0xb6f92e5f - sys_common::unwind::begin_unwind_inner::h1d39a3f983dfda740gs
   5: 0xb6f803df - sys_common::unwind::begin_unwind::begin_unwind::h3512061288630715681
   6: 0xb6f83d63 - mapper::create_mapper::he15687723fbb39d4uVd
   7: 0xb6f86717 - start_emulator::ha09827b57096d9acNGf
   8: 0xb6f7e437 - main::h4c1a5b6fa5311b62Qda

Compiled with 1.5, run at 16bpp colour depth. I think README.md might need more info on requirements :)

Error compiling cpu.rs:

$ make
rustc --cfg ndebug --cfg ncpuspew -O -L . nes.rc -o nes || echo ""If this build failed due to missing SDL bindings, please install them from https://github.com/brson/rust-sdl and copy the .dll/.dylib/.so into this directory or use RUSTLDFLAGS.""
cpu.rs:709:19: 709:21 error: No rules expected the token: op
cpu.rs:709         decode_op!(op: op, this: self);
                              ^~

My rustc was compiled from mozilla/rust@917d3c28b6aab7de454087236173122616a6f9b6.

It's a syntax update (replaced static with const for global constants, .. with ... for ranges in match, etc.). Also updated the calls to the rust-sdl2 library to bring it inline with the newest version. Finally, I added a .gitignore to ignore the target directory and Cargo.lock when add files to the staging area.

Now builds and runs, probably slower than before, but playable with -O compile flag.

Only testing was an enjoyable run through 1-1 to 1-3 and playing with save states.

Trying to learn Rust by studying this codebase. It doesn't compile with the latest rustc, and I'm a so new to the language that I don't know how to fix this:

/rust/sprocketnes$ make rustc --cfg ndebug --cfg ncpuspew -O -L . nes.rc -o nes || echo ""If this build failed due to missing SDL bindings, please install the
m from https://github.com/brson/rust-sdl and copy the .dll/.dylib/.so into this directory or use RUSTLDFLAGS."" mem.rs:104:16: 104:22 error: cannot borrow immutable dereference of & pointer as mutable mem.rs:104 mapper.prg_loadb(addr) ^~~~~~ mem.rs:122:16: 122:22 error: cannot borrow immutable dereference of & pointer as mutable mem.rs:122 mapper.prg_storeb(addr, val) ^~~~~~ ppu.rs:181:16: 181:22 error: cannot borrow immutable dereference of & pointer as mutable ppu.rs:181 mapper.chr_loadb(addr) ^~~~~~ ppu.rs:195:16: 195:22 error: cannot borrow immutable dereference of & pointer as mutable ppu.rs:195 mapper.chr_storeb(addr, val) ^~~~~~ ppu.rs:773:19: 773:25 error: cannot borrow immutable dereference of & pointer as mutable ppu.rs:773 if mapper.next_scanline() == Irq { ^~~~~~ error: aborting due to 5 previous errors If this build failed due to missing SDL bindings, please install them from https://github.com/brson/rust-sdl and copy the .dll/.dylib/
.so into this directory or use RUSTLDFLAGS.

/rust/sprocketnes$ rustc --version rustc 0.8-pre (bb35e23 2013-08-30 21:40:32 -0700) host: x86_64-unknown-linux-gnu

I did fix a one earlier compile error as follows:

/rust/sprocketnes$ git branch -v

• master 8b6fcfa Update for language changes and go runtimeless!

/rust/sprocketnes$ git diff diff --git a/util.rs b/util.rs^[[m index ac46941..dece364 100644^[[m --- a/util.rs^[[m +++ b/util.rs^[[m @@ -10,6 +10,7 @@ use std::libc::time_t;^[[m use std::libc;^[[m use std::ptr::null;^[[m use std::uint;^[[m +use std::c_str::ToCStr;^[[m ^[[m //^[[m // Standard library I/O replacements^[[m @@ -37,7 +38,7 @@ impl Fd {^[[m ForReading => O_RDONLY,^[[m ForWriting => O_WRONLY | O_CREAT | O_TRUNC^[[m } as c_int;^[[m

•        do path.as_c_str |c_path| {^[[m
  

•        do path.to_c_str().with_ref |c_path| {^[[m
           Fd {^[[m
               contents: libc::open(c_path, fd_mode, 493),^[[m
           }^[[m
  

Trying to compile (using make) on osx lion 10.8.2 gives

    rustc --cfg ndebug --cfg ncpuspew -O -L . nes.rc -o nes || echo ""If this build failed due to missing SDL bindings, please install them from https://github.com/brson/rust-sdl and copy the .dll/.dylib/.so into this directory or use RUSTLDFLAGS.""
    cpu.rs:338:16: 338:19 error: expected `{` but found `for`
    cpu.rs:338 impl<M:Mem> Mem for Cpu<M> {
                               ^~~
    If this build failed due to missing SDL bindings, please install them from https://github.com/brson/rust-sdl and copy the .dll/.dylib/.so into this directory or use RUSTLDFLAGS.

There is an error about needing the sdl dylib, but I have it in the directory.

So far I've been finding these by playing a game until it crashes. To find all of them, I'd probably need to do a thorough line-by-line audit or run a fuzzer.

I might try using afl-fuzz to catch the rest of these, sometime.

* Rename fail! to panic!
* Remove occurrences of "use std::owned::Box".
* Renderer<T> is now just Renderer in rust-sdl2.
* Rename lint from "ctypes" to "improper_ctypes".

https://github.com/pcwalton/sprocketnes/blob/163e5da03c6c1db2f78898a001f13d4dd739f0ff/src/disasm.rs#L93

format!() will already return a String therefore the call to .to_string() is meaningless.

This is achieved by eliminating two global variables and making the code work nicely with ownership analysis.

This is possible because AudioDevice has a lock method that returns an object that dereferences to NesCallback we can store the output buffer directly in NesCallback and then safely access it to update the raw sound data.

The resulting code is easier to follow and contains fewer unsafe blocks.

Apparently the APU is the bottleneck. Because when you disable it the emulator is blazingly fast (way too fast in fact). I'll see if I can optimize it.

    Updating registry `https://github.com/rust-lang/crates.io-index`
 Downloading libc v0.2.4
 Downloading lazy_static v0.1.15
 Downloading sdl2 v0.6.1
 Downloading time v0.1.34
 Downloading bitflags v0.2.1
 Downloading rand v0.3.12
 Downloading num v0.1.29
 Downloading libc v0.1.12
 Downloading sdl2-sys v0.6.2
 Downloading advapi32-sys v0.1.2
 Downloading winapi v0.2.5
 Downloading winapi-build v0.1.1
 Downloading rustc-serialize v0.3.16
 Downloading kernel32-sys v0.2.1
   Compiling winapi-build v0.1.1
   Compiling winapi v0.2.5
   Compiling libc v0.2.4
   Compiling sdl2-sys v0.6.2
   Compiling rustc-serialize v0.3.16
   Compiling libc v0.1.12
   Compiling bitflags v0.2.1
   Compiling lazy_static v0.1.15
error[E0642]: patterns aren't allowed in methods without bodies
   --> /Users/me/.cargo/registry/src/github.com-1ecc6299db9ec823/rustc-serialize-0.3.16/src/serialize.rs:145:45
    |
145 |                                             &f_name: &str,
    |                                             ^^^^^^^

   Compiling advapi32-sys v0.1.2
   Compiling kernel32-sys v0.2.1
error: aborting due to previous error

For more information about this error, try `rustc --explain E0642`.
error: Could not compile `rustc-serialize`.
warning: build failed, waiting for other jobs to finish...
error: build failed


➜  sprocketnes git:(master) ✗ rustc --explain E0642
error: no extended information for E0642```

For example, here's how "inc" is implemented:

// Increments and decrements
fn inc<AM:AddressingMode<M>>(&mut self, am: AM) {
    let val = am.load(self);
    let val = self.set_zn(val + 1);
    am.store(self, val)
}

This is "set_zn":

 fn set_zn(&mut self, val: u8) -> u8 {
    self.set_flag(ZERO_FLAG, val == 0);
    self.set_flag(NEGATIVE_FLAG, (val & 0x80) != 0);
    val
}

I can't see how this would handle a increment of value 255 without panicking? Seeing that there's been no commits in 3 years, perhaps Rust handled overflows differently back then? (Just getting started, writing an emulator in Rust to get to know the language)