sexpr-wasm

Translates from WebAssembly s-expressions to v8-native-prototype bytecode.

Cloning

Clone as normal, but don't forget to update/init submodules as well:

$ git clone https://github.com/WebAssembly/sexpr-wasm-prototype
$ git submodule update --init

This will fetch the v8-native-prototype repo, which is needed for some tests.

Building

Building just sexpr-wasm:

$ make
mkdir out
cc -Wall -Werror -g -c -o out/sexpr-wasm.o src/sexpr-wasm.c
cc -Wall -Werror -g -c -o out/wasm-parse.o src/wasm-parse.c
cc -Wall -Werror -g -c -o out/wasm-gen.o src/wasm-gen.c
cc -o out/sexpr-wasm out/sexpr-wasm.o out/wasm-parse.o out/wasm-gen.o

If you make changes to src/hash.txt, you'll need to install gperf as well. On Debian-based systems:

$ sudo apt-get install gperf
...
$ touch src/hash.txt
$ make
gperf --compare-strncmp --readonly-tables --struct-type src/hash.txt --output-file src/hash.h
...

Building v8-native-prototype

The v8-native-prototype lets you run the generated bytecode. Some of the tests rely on this repo. To build it:

$ scripts/build-d8.sh
...

When it is finished, there will be a d8 executable in the third_party/v8-native-prototype/v8/v8/out/Release directory.

Running

First write some WebAssembly s-expressions:

$ cat > test.wasm << HERE
(module
  (export "test" 0)
  (func (result i32)
    (i32.add (i32.const 1) (i32.const 2))))
HERE

Then run sexpr-wasm to build v8-native-prototype bytecode:

$ out/sexpr-wasm test.wasm -o test.bin

This can be loaded into d8 using JavaScript like this:

$ third_party/v8-native-prototype/v8/v8/out/Release/d8
V8 version 4.7.0 (candidate)
d8> buffer = readbuffer('test.bin');
[object ArrayBuffer]
d8> module = WASM.instantiateModule(buffer, {});
{memory: [object ArrayBuffer], test: function test() { [native code] }}
d8> module.test()
3

If you just want to run a quick test, you can use the run-d8.py script instead:

$ test/run-d8.py test.wasm
test() = 3

Tests

To run tests:

$ make test
[+251|-0|%100] (0.55s)

In this case, there were 251 passed tests and no failed tests, which took .55 seconds to run.

You can also run the Python test runner script directly:

$ test/run-tests.py
[+251|-0|%100] (0.40s)

$ test/run-tests.py -v
+ bad-string-escape.txt (0.002s)
+ bad-string-hex-escape.txt (0.004s)
+ bad-string-eof.txt (0.005s)
+ bad-toplevel.txt (0.003s)
...

To run a subset of the tests, use a glob-like syntax:

$ test/run-tests.py const -v
+ dump/const.txt (0.003s)
+ expr/bad-const-i32-garbage.txt (0.002s)
+ expr/bad-const-f32-trailing.txt (0.004s)
+ expr/bad-const-i32-overflow.txt (0.003s)
+ expr/bad-const-i32-trailing.txt (0.002s)
+ expr/bad-const-i32-just-negative-sign.txt (0.003s)
+ expr/const.txt (0.003s)
+ expr/bad-const-i32-underflow.txt (0.004s)
+ expr/bad-const-i64-overflow.txt (0.005s)
[+9|-0|%100] (0.02s)

$ test/run-tests.py expr*const*i32 -v
+ expr/bad-const-i32-garbage.txt (0.004s)
+ expr/bad-const-i32-overflow.txt (0.002s)
+ expr/bad-const-i32-trailing.txt (0.002s)
+ expr/bad-const-i32-just-negative-sign.txt (0.002s)
+ expr/bad-const-i32-underflow.txt (0.002s)
[+5|-0|%100] (0.01s)