Add import section to modules #54
Conversation
| @@ -172,13 +168,13 @@ expr: | |||
| ( if <expr> <expr> ) ;; = (if <expr> <expr> (nop)) | |||
| ( loop <expr>* ) ;; = (loop (block <expr>*)) | |||
| ( label <name>? <expr>* ) ;; = (label (block <expr>*)) | |||
| ( break <var> <expr>* ) | |||
| ( break <var> <expr> ) | |||
There was a problem hiding this comment.
shouldn't this just be ( break ) now? I believe the expression list was used as a return value to the label.
There was a problem hiding this comment.
You're right there a bug here, but, because labels still return a value (this patches changes that from value list to value option) this should be (break <var> <expr>?).
|
LGTM |
|
Thanks for reviewing. I'm inclined to wait for @rossberg-chromium to return before merging, unless he was going to be offline for a while. Anyone know? |
|
@rossberg-chromium Updated |
|
Could you do the corresponding change in AstSemantics before merging? The idea sounds fine to me, I just want to make sure the more-oft-observed design repo gets a say in this! |
|
@jfbastien I was about to say that this PR simply implements what's already in |
|
LGTM |
Add import section to modules
Revert "Check for invalid extend operation (WebAssembly#64)" This reverts commit eb056c8. Revert "Implement extend instructions in interpreter (WebAssembly#60)" This reverts commit 2f7592b. Revert "Fix naming of `iextend` macro in math.def (WebAssembly#56)" This reverts commit ae08b5e. Revert "Add spec text for extend operators (WebAssembly#54)" This reverts commit a4cb852.
This adds the new operations, as well as the new data and element segment formats. Implemented: * Text parsing/writing * Binary encoding/decoding * Validation TODO: * Tests * DataCount section * Proper encoding of passive element segments (using ref.null etc) * Evaluation
Stack types are contravariant in all their parameters, including the return stack reference.
Merge with upstreams
This patch builds on #53 and implements imports of builtin functions as discussed in #47. The main difference is that, instead of injecting imports into the index space of functions (such that, after importing a function you call it with
(call ...), the patch creates a separate index space for imports that are called through a new(call_import ...). The reason for this is that I think it makes everything (codegen, spec, impl) simpler.