Limits

[SoniEx2]

wabt currently applies no limits anywhere ever.

we would like to introduce a limits.json file that can be passed in the CLI. (we think a json would be a better fit than putting it all in the CLI.) like --limits limits.json.

the hard part is enforcing the limits tbh... but anyway.

[keithw]

Hmm, I'm not sure I totally follow. The BinaryReaderIR does impose limits like kMaxNestingDepth, kMaxFunctionLocals, kMaxFunctionParams, and kMaxFunctionResults, which match the values in V8 (https://chromium.googlesource.com/v8/v8/+/refs/heads/main/src/wasm/wasm-limits.h) and hopefully the other big engines.

You're saying WABT should limit other kinds of things? And/or the values should be configurable at runtime? I think prevailing practice so far has just been to sort of handshake agree on reasonable values for the various "implementation limitations" (https://webassembly.github.io/spec/core/appendix/implementation.html) across major implementations.

[SoniEx2]

right, so

wabt/src/binary-reader-ir.cc

Lines 94 to 97 in ea193b4

	static constexpr size_t kMaxNestingDepth = 16384; // max depth of label stack
	static constexpr size_t kMaxFunctionLocals = 50000; // matches V8
	static constexpr size_t kMaxFunctionParams = 1000; // matches V8
	static constexpr size_t kMaxFunctionResults = 1000; // matches V8

... we would never have found these if you hadn't mentioned them. (... why aren't they in a header??? or, alternatively, configurable by the user.)

it is worth noting that these only apply to the IR reader, they do not apply to text format or - more importantly - to interpreter.

[sbc100]

Making them configurable seems like maybe unnecessary. Putting them in a header seems pretty reasonable. Maybe just in the existing header for the binary file stuff.

[SoniEx2]

we feel like making them configurable might be useful given the whole wabt is primarily a developer tool (tho having defaults probably wouldn't hurt), but we mean... maybe. we also need to get these to apply to the interpreter and the text format and whatnot tho.

[sbc100]

Hopefully only the parsers / validators need to enforce these limits. The interpreter can assume all its inputs have been validated. There may be some exceptions but I can't think of any

[SoniEx2]

hmm... we'll think about it.