[Parser] Parse wast scripts #6581
Conversation
The spec tests use an extension of the standard text format that includes various commands and assertions used to test WebAssembly implementations. Add a utility to parse this extended WebAssembly script format and use it in wasm-shell to check that it parses our spec tests without error. Fix a few errors the new parser found in our spec tests. A future PR will rewrite wasm-shell to interpret the results of the new parser, but for now to keep the diff smaller, do not do anything with the new parser except check for errors.
|
This stack of pull requests is managed by Graphite. Learn more about stacking. |
| ;; (assert_return (invoke "tuple-global-get") (tuple.make 2 (i32.const 0) (i64.const 0))) | ||
| ;; (assert_return (invoke "tuple-global-set")) | ||
| ;; (assert_return (invoke "tuple-global-get") (tuple.make 2 (i32.const 42) (i64.const 7))) | ||
| ;; (assert_return (invoke "tail-call") (tuple.make 2 (i32.const 42) (i64.const 7))) |
There was a problem hiding this comment.
Are these not valid? (If so why not remove them?)
There was a problem hiding this comment.
The wast parser does not parse tuple.make.
There was a problem hiding this comment.
Naively I'd expect the wast parser to parse a superset of the wat parser's input, which I guess is related to the other question about sharing logic between them?
There was a problem hiding this comment.
Since the goal is to run the upstream spec tests, it shouldn't be necessary to parse tuple.make to test multivalue. Rather than doing a bunch of work to support parsing tuple.make constants anyway, it seemed better to comment this out for now. I didn't delete them entirely because when we switch to the upstream spec tests, we will want to check that we regain this coverage.
There was a problem hiding this comment.
Fair enough. I'd have hoped to see tuple.make get parsed properly by reusing some wat parser logic, but as that's not the goal here it doesn't matter.
src/parser/wast-parser.cpp
Outdated
| return in.err("expected end of v128.const"); | ||
| } | ||
| return vec; | ||
| } |
There was a problem hiding this comment.
There seems to be duplication here of the core parser logic for these constants etc. - can that be shared somehow?
There was a problem hiding this comment.
I initially implemented this by calling parseExpression, then translating the parsed constant Expression* into a literal. I forget why reimplementing the parsing ended up being easier, so I'll try that again.
There was a problem hiding this comment.
Ok, this reuses the existing parser now. PTAL.
| ;; (assert_return (invoke "tuple-global-get") (tuple.make 2 (i32.const 0) (i64.const 0))) | ||
| ;; (assert_return (invoke "tuple-global-set")) | ||
| ;; (assert_return (invoke "tuple-global-get") (tuple.make 2 (i32.const 42) (i64.const 7))) | ||
| ;; (assert_return (invoke "tail-call") (tuple.make 2 (i32.const 42) (i64.const 7))) |
There was a problem hiding this comment.
Fair enough. I'd have hoped to see tuple.make get parsed properly by reusing some wat parser logic, but as that's not the goal here it doesn't matter.
The spec tests use an extension of the standard text format that includes
various commands and assertions used to test WebAssembly implementations. Add a
utility to parse this extended WebAssembly script format and use it in
wasm-shell to check that it parses our spec tests without error. Fix a few
errors the new parser found in our spec tests.
A future PR will rewrite wasm-shell to interpret the results of the new parser,
but for now to keep the diff smaller, do not do anything with the new parser
except check for errors.