port to zig 0.16.0

[dasimmet]

seems to work for me so far, but i had to vendor the new standard library std.Io.Reader.takeLeb128
to make it not read extra bytes to pass the test suite.
also error enums returned from std.Io.Reader.takeEnum changed to error.InvalidEnumTag and i had to adapt testsuite errors to match.

[malcolmstill]

Thanks for this @dasimmet.

Are you able to explain the LEB128 issue a bit more as you understand it? Would be great not to have to vendor the code so I'm keen to understand if there are any alternatives here before committing to that.

[dasimmet]

The new implementation of 0.16 takeLeb will happily keep reading bytes from the reader as long as the continuation bit is set due to the final while loop, even if its more than 5 bytes (for 32bit ints):

22dc32f#diff-2bc993f98f63c8cc2df8f9246692eae522dad0cdda50e48e692803979a04662bL95

The wasm spec forbids such "overlong" leb128 values, while technically i think they are valid. Thats what these "integer representation too long" cases test.

I guess we can search or file an issue with zig to see if std behavior is intentional, or check after calling takeleb128 if more bytes were consumed maybe

[malcolmstill]

Thanks @dasimmet