Same test expected to result in different error messages

[csjh]

spec/test/core/data.wast

Line 298 in 1904486

(assert_invalid

spec/test/core/data.wast

Line 319 in 1904486

(assert_invalid

The above 2 tests seem to generate the same bytes, but two different error messages are expected from them. Not too sure which to interpret as the correct one.

[keithw]

I believe the spec interpreter only requires the expected error message to be a prefix of its actual error message, and in #1076 and #1089, the expected error messages at the time were edited to remove the indices (so that implementations wouldn't be required to produce a dynamically formatted error message). Some later tests have come in after this...

[csjh]

Interesting - are some of these test quirks documented anywhere? There's a few I've picked up on (spectest globals initializing with the value 666 comes to mind) which are a bit hard to pick up on w/o asking around

[rossberg]

Unfortunately, not really. The closest you get to documentation for the .wast format is here. At least the signature of the spectest module is listed there as well, though not its values.

I'll add that information to that doc.

[rossberg]

Done.

[csjh]

Thanks!