Boris Zbarsky [:bzbarsky] Assignee
	Description • 12 years ago

The current setup doesn't really scale very well, now that I've tried to convert CSSStyleDeclaration and the webgl context.

	Boris Zbarsky [:bzbarsky] Assignee
	Comment 2 • 12 years ago

Comment on attachment 626337 [details] [diff] [review]
Move infallibility annotations into webidl.   this new setup, there are three new extended attributes: Infallible,

Peter, what do you think?  Both about the overall setup and the particular names I picked?

In some ways it would be nice to have a syntax that we could put after the attribute to avoid changing the line with the per-spec extended attributes, but I don't really want to hack up our parser if I don't have to....

I left the old mechanism in for now, but I can remove it if we think this works fine.

I would also be ok with nixing the worker/mainthread stuff and either requiring that things be flagged as infallible only if they're infallible on both workers and mainthread or that people who want to specify that sort of thing need to do so via the conf file...  But I think this setup is actually pretty simple.

	Peter Van der Beken [:peterv]
	Comment 3 • 12 years ago

Comment on attachment 626337 [details] [diff] [review]
Move infallibility annotations into webidl.   this new setup, there are three new extended attributes: Infallible,

Review of attachment 626337 [details] [diff] [review]:
-----------------------------------------------------------------

::: dom/bindings/Configuration.py
@@ +179,5 @@
>      def getExtendedAttributes(self, member, getter=False, setter=False):
>          name = member.identifier.name
>          if member.isMethod():
> +            attrs = self.extendedAttributes['all'].get(name, [])
> +            # Constructors seem to not have useful extended attr dicts?

What's this about? If that's why we have the |try| below I'd rather fix constructors. Aren't constructors IDLMethods?

@@ +188,5 @@
> +                if (infallible is not None and
> +                    (infallible is True or
> +                     ('Workers' in infallible and self.workers) or
> +                     ('MainThread' in infallible and not self.workers))):
> +                    attrs.append("infallible")

Do we want to throw for other values?

@@ +207,5 @@
> +        if (infallible is not None and
> +            (infallible is True or
> +             ('Workers' in infallible and self.workers) or
> +             ('MainThread' in infallible and not self.workers))):
> +            attrs.append("infallible")

Same here?

	Boris Zbarsky [:bzbarsky] Assignee
	Comment 4 • 12 years ago

> If that's why we have the |try| below I'd rather fix constructors. Aren't constructors
> IDLMethods?

That's what it's about, yes.  Constructors are IDLMethods, but IDLInterfaceMember only sets up self._extendedAttrDict when addExtendedAttributes is called, and it's not called for constructors.

I'll just move that to IDLInterfaceMember.__init__.

> Do we want to throw for other values?

I can do that, yes.  Both places.

	Boris Zbarsky [:bzbarsky] Assignee
	Comment 5 • 12 years ago

https://hg.mozilla.org/integration/mozilla-inbound/rev/ddfa6d5ad10a

	Ryan VanderMeulen [:RyanVM]
	Comment 6 • 12 years ago

https://hg.mozilla.org/mozilla-central/rev/ddfa6d5ad10a