sexpr-wasm-prototype

A suite of build tools for WebAssembly, including:

• sexpr-wasm: translate from s-expressions to the WebAssembly binary-encoding
• wasm-wast: the inverse of sexpr-wasm, translate from the binary encoding back to an s-expression source file (also known as a .wast)
• wasm-interp: decode and run a WebAssembly binary file using a stack-based interpreter

Cloning

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

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

This will fetch the testsuite and gtest repos, which are needed for some tests.

Building

You'll need CMake. If you just run make, it will run CMake for you, and put the result in out/clang/Debug/ by default:

$ make

This will build the default version of the tools: a debug build using the Clang compiler. It will also create a symlinks to the built binaries in the out/ directory.

There are many make targets available for other configurations as well. They are generated from every combination of a compiler, build type and configuration.

• compilers: gcc, clang, gcc-i686, gcc-fuzz
• build types: debug, release
• configurations: empty, asan, msan, lsan, ubsan, no-re2c-bison, no-tests

They are combined with dashes, for example:

$ make clang-debug
$ make gcc-i686-release
$ make clang-debug-lsan
$ make gcc-debug-no-re2c-bison

You can also run CMake yourself, the normal way:

$ mkdir build
$ cd build
$ cmake ..
...

If you make changes to src/wasm-ast-parser.y, you'll need to install Bison. Before you upload your PR, please run make update-bison to update the prebuilt C sources in src/prebuilt/.

If you make changes to src/wasm-ast-lexer.c, you'll need to install re2c. Before you upload your PR, please run make update-re2c to update the prebuilt C sources in src/prebuilt/.

CMake will detect if you don't have re2c or Bison installed and use the prebuilt source files instead.

Running sexpr-wasm

Some examples:

# parse and typecheck test.wast
$ out/sexpr-wasm test.wast

# parse test.wast and write to binary file test.wasm
$ out/sexpr-wasm test.wast -o test.wasm

# parse spec-test.wast, and write verbose output to stdout (including the
# meaning of every byte)
$ out/sexpr-wasm spec-test.wast -v

# parse spec-test.wast, and write files to spec-test.json. Modules are written
# to spec-test.0.wasm, spec-test.1.wasm, etc.
$ out/sexpr-wasm spec-test.wast --spec -o spec-test.json

You can use -h to get additional help:

$ out/sexpr-wasm -h

Or try the online demo.

Running wasm-wast

Some examples:

# parse binary file test.wasm and write s-expression file test.wast
$ out/wasm-wast test.wasm -o test.wast

# parse test.wasm and write test.wast, using the debug names, if any
$ out/wasm-wast test.wasm --debug-names -o test.wast

You can use -h to get additional help:

$ out/wasm-wast -h

Running wasm-interp

Some examples:

# parse binary file test.wasm, and type-check it
$ out/wasm-interp test.wasm

# parse test.wasm and run all its exported functions
$ out/wasm-interp test.wasm --run-all-exports

# parse test.wasm, run the exported functions and trace the output
$ out/wasm-interp test.wasm --run-all-exports --trace

# parse test.json and run the spec tests
$ out/wasm-interp test.json --spec

# parse test.wasm and run all its exported functions, setting the value stack
# size to 100 elements
$ out/wasm-interp test.wasm -V 100 --run-all-exports

As a convenience, you can use test/run-interp.py to convert a .wast file to binary first, then run it in the interpreter:

$ test/run-interp.py --spec spec-test.wast
20/20 tests.passed.

You can use -h to get additional help:

$ out/wasm-interp -h
$ out/run-interp.py -h

Running the test suite

To run all the tests with default configuration:

$ make test

Every make target has a matching test-* target.

$ make gcc-debug-asan
$ make test-gcc-debug-asan
$ make clang-release
$ make test-clang-release
...

You can also run the Python test runner script directly:

$ test/run-tests.py

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

$ test/run-tests.py const -v
+ dump/const.txt (0.002s)
+ parse/assert/bad-assertreturn-non-const.txt (0.003s)
+ parse/expr/bad-const-i32-overflow.txt (0.002s)
+ parse/expr/bad-const-f32-trailing.txt (0.004s)
+ parse/expr/bad-const-i32-garbage.txt (0.005s)
+ parse/expr/bad-const-i32-trailing.txt (0.003s)
+ parse/expr/bad-const-i32-underflow.txt (0.003s)
+ parse/expr/bad-const-i64-overflow.txt (0.002s)
+ parse/expr/bad-const-i32-just-negative-sign.txt (0.004s)
+ parse/expr/const.txt (0.002s)
[+10|-0|%100] (0.11s)

$ test/run-tests.py expr*const*i32
+ parse/expr/bad-const-i32-just-negative-sign.txt (0.002s)
+ parse/expr/bad-const-i32-overflow.txt (0.003s)
+ parse/expr/bad-const-i32-underflow.txt (0.002s)
+ parse/expr/bad-const-i32-garbage.txt (0.004s)
+ parse/expr/bad-const-i32-trailing.txt (0.002s)
[+5|-0|%100] (0.11s)

When tests are broken, they will give you the expected stdout/stderr as a diff:

$ <whoops, turned addition into subtraction in the interpreter>
$ test/run-tests.py interp/binary
- interp/binary.txt
  STDOUT MISMATCH:
  --- expected
  +++ actual
  @@ -1,4 +1,4 @@
  -i32_add() => i32:3
  +i32_add() => i32:4294967295
   i32_sub() => i32:16
   i32_mul() => i32:21
   i32_div_s() => i32:4294967294

**** FAILED ******************************************************************
- interp/binary.txt
[+0|-1|%100] (0.13s)

Writing New Tests

Tests must be placed in the test/ directory, and must have the extension .txt. The directory structure is mostly for convenience, so for example you can type test/run-tests.py interp to run all the interpreter tests. There's otherwise no logic attached to a test being in a given directory.

That being said, try to make the test names self explanatory, and try to test only one thing. Also make sure that tests that are expected to fail start with bad-.

The test format is straightforward:

;;; KEY1: VALUE1A VALUE1B...
;;; KEY2: VALUE2A VALUE2B...
(input (to)
  (the executable))
(;; STDOUT ;;;
expected stdout
;;; STDOUT ;;)
(;; STDERR ;;;
expected stderr
;;; STDERR ;;)

The test runner will copy the input to a temporary file and pass it as an argument to the executable (which by default is out/sexpr-wasm).

The currently supported list of keys:

• TOOL: a set of preconfigured keys, see below.
• EXE: the executable to run, defaults to out/sexpr-wasm
• STDIN_FILE: the file to use for STDIN instead of the contents of this file.
• FLAGS: additional flags to pass to the executable
• ERROR: the expected return value from the executable, defaults to 0
• SLOW: if defined, this test's timeout is doubled.
• SKIP: if defined, this test is not run. You can use the value as a comment.
• TODO,NOTE: useful place to put additional info about the test.

The currently supported list of tools:

• sexpr-wasm: runs sexpr-wasm
• run-roundtrip: runs the run-roundtrip.py script. This does a roundtrip conversion using sexpr-wasm and wasm-wast, making sure the .wasm results are identical.
• run-interp: runs the run-interp.py script, running all exported functions
• run-interp-spec: runs the run-interp.py script with the --spec flag

When you first write a test, it's easiest if you omit the expected stdout and stderr. You can have the test harness fill it in for you automatically. First let's write our test:

$ cat > test/my-awesome-test.txt << HERE
;;; TOOL: run-interp-spec
(module
  (export "add2" 0)
  (func (param i32) (result i32)
    (i32.add (get_local 0) (i32.const 2))))
(assert_return (invoke "add2" (i32.const 4)) (i32.const 6))
(assert_return (invoke "add2" (i32.const -2)) (i32.const 0))
HERE

If we run it, it will fail:

- my-awesome-test.txt
  STDOUT MISMATCH:
  --- expected
  +++ actual
  @@ -0,0 +1 @@
  +2/2 tests passed.

**** FAILED ******************************************************************
- my-awesome-test.txt
[+0|-1|%100] (0.03s)

We can rebase it automatically with the -r flag. Running the test again shows that the expected stdout has been added:

$ test/run-tests.py my-awesome-test -r
[+1|-0|%100] (0.03s)
$ test/run-tests.py my-awesome-test
[+1|-0|%100] (0.03s)
$ tail -n 3 test/my-awesome-test.txt
(;; STDOUT ;;;
2/2 tests passed.
;;; STDOUT ;;)

Sanitizers

To build with the LLVM sanitizers, append the sanitizer name to the target:

$ make clang-debug-asan
$ make clang-debug-msan
$ make clang-debug-lsan
$ make clang-debug-ubsan

There are configurations for the Address Sanitizer (ASAN), Memory Sanitizer (MSAN), Leak Sanitizer (LSAN) and Undefine Behavior Sanitizer (UBSAN). You can read about the behaviors of the sanitizers in the link above, but essentially the Address Sanitizer finds invalid memory accesses (use after free, access out-of-bounds, etc.), Memory Sanitizer finds uses of uninitialized memory, the Leak Sanitizer finds memory leaks, and the Undefined Behavior Sanitizer finds undefined behavior (surprise!).

Typically, you'll just want to run all the tests for a given sanitizer:

$ make test-asan

You can also run the tests for a release build:

$ make test-clang-release-asan
...

The Travis bots run all of these tests (and more). Before you land a change, you should run them too. One easy way is to use the test-everything target:

$ make test-everything

To run everything the Travis bots do, you can use the following scripts:

$ CC=gcc scripts/travis-build.sh
$ CC=gcc scripts/travis-test.sh
$ CC=clang scripts/travis-build.sh
$ CC=clang scripts/travis-test.sh