WASM `write` creates slice with null pointer

[GnomedDev]

Describe the bug
The WASM write function is called with null pointer via the JS wrapper, which causes a panic when compiled in debug mode and UB in release mode.

sax-wasm/src/js/saxWasm.ts

Lines 674 to 675 in e4ea343

	this.writeBuffer.set(chunk, 0);
	write(0, chunk.byteLength);

sax-wasm/src/sax_wasm.rs

Lines 48 to 49 in e4ea343

	pub unsafe extern "C" fn write(ptr: *const u8, length: usize) {
	let document = slice::from_raw_parts(ptr, length);

To Reproduce

1. Compile with a recent stable compiler without --release.
2. Call SAXParser.write.
3. WASM crashes with unreachable instruction.

Expected behavior
The Rust code does not construct a null slice.

Desktop (please complete the following information):

• OS: Linux
• Browser: Firefox 135, Chromium 133.0.6943.98

Additional context
I am trying to use this library to parse XML files that are nearly 1gb, and running into an infinite loop. I found this while trying to debug that.

[justinwilaby]

👀

[GnomedDev]

I suspect this also fixes #108 but I was not able to repro this. - #109

Have you tried compiling in debug mode, on the latest stable version? The crash will only occur if you use debug mode as otherwise it's just silent UB, but following the https://doc.rust-lang.org/stable/std/slice/fn.from_raw_parts.html docs:

data must be non-null, valid for reads for len * mem::size_of::() many bytes, and it must be properly aligned

[justinwilaby]

Yes, that's correct. I build without the --release flag and use the 1gb xml in the benchmark.mjs with just one iteration. No crash with #110 applied.

[justinwilaby]

@GnomedDev - This also seems to have been fixed by #110. Can you confirm?

[GnomedDev]

The debug assertion seems to not being triggered anymore, but from what I see it's still creating a slice with the pointer value 0 (null), which is still UB.

[justinwilaby]

Since I compile the FFIs with a wasm32 target and zero is a valid memory address in WASM, I'm hesitant to "fix" this but I understand the desire for strict adherence to Rust's rules.

Advancing the offset by 1 byte on the JS side is inconsequential. I'll have a PR up shortly.