1020869 - Latin1 strings: support IdToTypeId, StringIsArrayIndex etc

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Comment 1

[Jan de Mooij [:jandem]]

Attachment: Patch

I also had to fix PutEscapedStringImpl, the inference spew calls it in debug builds. We should probably stop doing that if the spew is disabled though...

Comment 2

[Brian Hackett [Laid off!]]

Comment on attachment 8434827 [details] [diff] [review]
Patch

Review of attachment 8434827 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/src/jsinferinlines.h
@@ +163,5 @@
>  }
>  
> +template <class Range>
> +inline jsid
> +AtomToTypeId(Range cp, jsid id)

I think this would be clearer if the id parameter were removed, the return type changed to bool and the name was IdIsNumericTypeId.

@@ +198,5 @@
>  
>      /*
>       * Check for numeric strings, as in js_StringIsIndex, but allow negative
>       * and overflowing integers.
>       */

This comment should go in the above new function.

::: js/src/vm/String.h
@@ +710,5 @@
>      bool isFlat() const MOZ_DELETE;
>      JSFlatString &asFlat() const MOZ_DELETE;
>  
> +    template <typename CharT>
> +    bool isIndexSlow(const CharT *s, uint32_t *indexp) const;

Maybe make this static and pass in the length as well?  It just seems kind of weird to have a member function getting its own chars pointer passed in.

Comment 3

[Jan de Mooij [:jandem]]

Pushed with comments addressed:

https://hg.mozilla.org/integration/mozilla-inbound/rev/cd4218cd70f9

Comment 4

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/cd4218cd70f9