1135961 - Implement subclassing of DOM objects

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, defect)

Comment 1

[Boris Zbarsky [:bzbarsky]]

Attachment: WIP

This should all work, I think, except for preserving the wrapper when using a non-default proto.  In particular, I need to sort out the proxy story there, since apparently preserving the wrapper for proxies is ... complicated.

Comment 2

[Boris Zbarsky [:bzbarsky]]

Note to self: Use new XMLSerializer() as a fast constructor to test perf impact of this stuff.

Comment 3

[Boris Zbarsky [:bzbarsky]]

Attachment: WIP2

OK, so the preserving works now... .  New problem is that my asserts trigger in a debug buid, because the Promise ctor actually does the wrapping in Promise::Constructor, so the bindings don't get a chance to use the right proto.  Need to think a bit about how to handle that.

Comment 4

[Boris Zbarsky [:bzbarsky]]

Attachment: With a fix for an Xray SNAFU

Comment 5

[Boris Zbarsky [:bzbarsky]]

So I think the main options are to either hack codegen for Promise to pass in the desired proto for now, pending Promise moving into SpiderMonkey, or to change all DOM constructors to pass in the desired proto, in case the callee wraps eagerly.

I'm pretty happy doing the former, honestly.  Promise is a pretty special snowflake.  Peter, do you have a preference?

Comment 6

[Boris Zbarsky [:bzbarsky]]

Attachment: With constructor special-casing for Promise and MozAbortablePromise; the browser now starts without hitting any asserts

Comment 7

[Boris Zbarsky [:bzbarsky]]

Attachment: Small patch to test this stuff manually on top of the newtarget support

Comment 8

[Boris Zbarsky [:bzbarsky]]

Attachment: Small manual testcase

Comment 9

[Boris Zbarsky [:bzbarsky]]

Attachment: Implement subclassing of DOM objects

Comment 10

[Boris Zbarsky [:bzbarsky]]

Attachment: With some better comments in GetDesiredProto

Comment 11

[Boris Zbarsky [:bzbarsky]]

Attachment: With the rooting hazard in DoGetOrCreateDOMReflector hopefully fixed

Comment 12

[Boris Zbarsky [:bzbarsky]]

Attachment: And even compiling, with the GetWrapper changes

Comment 13

[Peter Van der Beken [:peterv]]

Comment on attachment 8639901 [details] [diff] [review]
And even compiling, with the GetWrapper changes

Review of attachment 8639901 [details] [diff] [review]:
-----------------------------------------------------------------

Thanks for the comments, they're really helpful.

::: dom/bindings/Codegen.py
@@ +3446,5 @@
> +        if self.descriptor.proxy:
> +            preserveWrapper = dedent(
> +                """
> +                // For DOM proxies, the only reliable way to preserve the wrapper
> +                // is to force creation of the expando object.

:-(

Comment 14

[Boris Zbarsky [:bzbarsky]]

> :-(

I filed bug 1189822.

Comment 15

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/ccf304bee50c

Comment 16

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/ccf304bee50c