Re: Call for Review: Web IDL Testing document

On 12/10/12 9:19 AM, Arthur Barstow wrote:
> * Are the assertions that are already defined (see Section 3 in the
> Testing document), valid?

A big caveat: Testing objects that are not actually using WebIDL is 
pointless for conformance criteria purposes.  At least in Gecko's case, 
some objects are using WebIDL already, some are not yet; a number of the 
ones the assertion tests suggest using so far are not.  So we may end up 
with test passes on the latter objects that have nothing to do with 
WebIDL.  It's worth checking with other UAs what their situation is when 
picking the sets of objects to test....

"Objects should have an internal prototype of Object.prototype unless 
otherwise noted" is not valid as written.  In particular, 
Object.getPrototypeOf(Element.prototype) == Node.prototype.  The only 
objects WebIDL really defines whose protos are Object.prototype should 
be interface objects for callback interfaces and prototype objects for 
interfaces which have no ancestor interface, I think.

"objects defined to be function objects have an internal prototype of 
Function.prototype" is fine as written, but I recommend testing a few 
interface objects as well, including at least one interface object for 
an interface with ancestors.  XMLHttpRequest and EventTarget are good 
candidates.

"JavaScript function returning a value for a void function is discarded" 
the note about handleEvent is false.  The return value of functions set 
as an event handler is used, but the return value of functions added via 
addEventListener is in fact ignored in UAs last I checked.

The rest look ok to me.

It might be good to try to split the spec into pieces and hand the 
pieces out to people to work on or something...

-Boris

Received on Monday, 10 December 2012 14:46:12 UTC