[JSC] Do not check isValid() in op_new_regexp

[Yusuke Suzuki]

If we have an invalid regexp, we should emit op_throw_xxx bytecode instead of op_new_regexp.

[Yusuke Suzuki]

Created attachment 329788 [details]
Patch

[Saam Barati]

Comment on attachment 329788 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=329788&action=review

r=me with debug build fix.

> Source/JavaScriptCore/bytecompiler/NodesCodegen.cpp:149
> +    const Identifier& message = generator.parserArena().identifierArena().makeIdentifier(generator.vm(), bitwise_cast<const LChar*>(messageCharacters), strlen(messageCharacters));

bitwise_case -> static_cast?

> JSTests/stress/regexp-syntax-error-invalid-flags.js:18
> +    return /Hello/cocoa;

I vote for adding a few more invalid regrex tests

[Yusuke Suzuki]

Comment on attachment 329788 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=329788&action=review

Thank you!

>> Source/JavaScriptCore/bytecompiler/NodesCodegen.cpp:149
>> +    const Identifier& message = generator.parserArena().identifierArena().makeIdentifier(generator.vm(), bitwise_cast<const LChar*>(messageCharacters), strlen(messageCharacters));
> 
> bitwise_case -> static_cast?

Since this is `const char*` to `const LChar*` cast, bitwise_cast is necessary. (LChar is uint8_t).

>> JSTests/stress/regexp-syntax-error-invalid-flags.js:18
>> +    return /Hello/cocoa;
> 
> I vote for adding a few more invalid regrex tests

The other invalid RegExp becomes early (parser phase) syntax error. We have relaxed rule for invalid flags, which raise runtime errors.

[Yusuke Suzuki]

Committed r226209: <https://trac.webkit.org/changeset/226209>

[Radar WebKit Bug Importer]

<rdar://problem/36169188>