Multiple issues with emscripten tests: typedarray2, regexp syntax, parseJSFunc about duktape

[svaarala]

There seem to be multiple issues related to recent Emscripten changes.

First, Duktape does not currently support typed arrays (ArrayBuffer, UInt8Array etc) and it seems that Emscripten now requires them by default. Setting EMCC_FAST_COMPILER=0 in the env is needed to avoid this issue.

Then there is another issue with the following regexp in Emscripten generated output:

var sourceRegex = /^function\s\(([^)]*)\)\s*{\s*([^*]*?)[\s;]*(?:return\s*(.*?)[;\s]*)?}$/;

The brace characters ('{' and '}') are not escaped as required. This can be fixed manually (I'll submit a pull request for Emscripten for this).

After fixing this one gets:

$ ./duk /tmp/duk-emcc-test.js
TypeError: invalid base reference for property read
  duk_hobject_props.c:1938
  parseJSFunc /tmp/duk-emcc-test.js:653
  anon /tmp/duk-emcc-test.js:661
  global /tmp/duk-emcc-test.js:714 preventsyield

This is caused by the generated code trying to parse the toString() output of a function, e.g.:

function empty() {/* source code */}

The (fixed) regexp does not match this (which seems incorrect) but does match e.g.:

function () {}

The E5 spec only requires that function toString() output be in valid syntax (E5 Section 15.3.4.2):

An implementation-dependent representation of the function is returned. This representation
has the syntax of a FunctionDeclaration. Note in particular that the use and placement of white
space, line terminators, and semicolons within the representation String is implementation-
dependent.

The offending function is:

var sourceRegex = /^function\s\(([^)]*)\)\s*{\s*([^*]*?)[\s;]*(?:return\s*(.*?)[;\s]*)?}$/;
function parseJSFunc(jsfunc) {
  // Match the body and the return value of a javascript function source
  var parsed = jsfunc.toString().match(sourceRegex).slice(1);
  return {arguments : parsed[0], body : parsed[1], returnValue: parsed[2]}
}

Just as a quick fix, this makes the hello world test work again:

var sourceRegex = /^function\s\(([^)]*)\)\s*\{\s*([^*]*?)[\s;]*(?:return\s*(.*?)[;\s]*)?\}$/;
function parseJSFunc(jsfunc) {
  // Match the body and the return value of a javascript function source
  var parsed = jsfunc.toString().match(sourceRegex);
  if (!parsed) { return {}; }
  parsed = parsed.slice(1);
  return {arguments : parsed[0], body : parsed[1], returnValue: parsed[2]}
}

[svaarala]

The above comment reactivates fix_emscripten.py and adds the necessary regexp fixes to Emscripten output to allow Hello World test to work again. I'll submit fixes to Emscripten soon.

[svaarala]

The util/fix_emscripten.py now fixes a few more issues, all related to Emscripten using a few regexps outside the strict E5 syntax. There's a Ditz issue for tracking that they're submitted to Emscripten closer to 1.0, I'll close this now.