1213056 - Change RTCStatsReport to be maplike.

Status: RESOLVED FIXED

(Core :: WebRTC, defect, P1)

Description

[Jan-Ivar Bruaroey [:jib] (needinfo? me)]

Spec change.

Should keep old way for backwards compatibility as well for a release or two, with warnings.

Comment 1

[Jan-Ivar Bruaroey [:jib] (needinfo? me)]

Attachment: MozReview Request: Bug 1213056 - change getStats to be maplike.

Review commit: https://reviewboard.mozilla.org/r/39017/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/39017/

Comment 2

[Jan-Ivar Bruaroey [:jib] (needinfo? me)]

Attachment: MozReview Request: Bug 1213056 - update tests to use maplike getStats.

Review commit: https://reviewboard.mozilla.org/r/39019/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/39019/

Comment 3

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 8728685 [details]
MozReview Request: Bug 1213056 - change getStats to be maplike.

https://reviewboard.mozilla.org/r/39017/#review35751

::: dom/media/PeerConnection.js:318
(Diff revision 1)
> +    Object.defineProperties(this.__DOM_IMPL__.wrappedJSObject, legacyProps);

This is exposing the chrome getter functions directly to content, isn't it?  Are you sure that's OK?  Please check with bholley.

::: dom/webidl/RTCStatsReport.webidl
(Diff revision 1)
> -  void forEach(RTCStatsReportCallback callbackFn, optional any thisArg);

The forEach/get/has methods in the chrome impl are dead code now, right?  If so, please remove them.

r=me modulo those issues, but please do check with bholley.

Comment 4

[Jan-Ivar Bruaroey [:jib] (needinfo? me)]

https://reviewboard.mozilla.org/r/39017/#review35751

> This is exposing the chrome getter functions directly to content, isn't it?  Are you sure that's OK?  Please check with bholley.

FWIW the old code did something similar, expose a chrome value (with configurable:false, writable:false), but yes I changed it to a getter.

> The forEach/get/has methods in the chrome impl are dead code now, right?  If so, please remove them.

I thought I did, did I miss something?

Thanks, will do, but if you could hit the "Ship It" button in mozReview that would be great (required by auto-land I think).

Comment 5

[Boris Zbarsky [:bzbarsky]]

> FWIW the old code did something similar, expose a chrome value (with configurable:false, writable:false)

No, it didn't.  The old code did:

292     Object.defineProperties(this.__DOM_IMPL__.wrappedJSObject, props);

where "props" is a (chrome-side, yes, but that doesn't matter in this case) object containing descriptors with "value: stat" where "stat" is the argument to forEach.  And forEach was implemented as:

297       cb.call(thisArg || this._report, this.get(key), key, this._report);

where this.get(key) returns like so:

307       let pubobj = Cu.createObjectIn(win);
308       Object.defineProperties(pubobj, props);
309       return pubobj;

so it was returning content-side objects and those were used as the values of the properties defined on the content-side reflector by makeStatsPublic.  Content never got to see chrome-side anything.

> I thought I did, did I miss something?

No, I did, because of the crappy diff viewer in mozreview.  :(

> but if you could hit the "Ship It" button in mozReview

Gah.  I thought I had.  How I hate mozreview....

Comment 6

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 8728685 [details]
MozReview Request: Bug 1213056 - change getStats to be maplike.

https://reviewboard.mozilla.org/r/39017/#review35783

Note that this still needs sign-off from bholley on the exposing chrome getters bit.  Of course I also couldn't redirect the original review request to him, because mozreview....

Comment 7

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 8728685 [details]
MozReview Request: Bug 1213056 - change getStats to be maplike.

https://reviewboard.mozilla.org/r/39017/#review35789

::: dom/media/PeerConnection.js:309
(Diff revision 1)
> +        get: function() {

OK, I checked with bholley.  We do not, in fact, want to expose the chrome function directly.  Use Cu.cloneInto with cloneFunctions=true to clone things involving accessors into the content compartment.

This time I'm removing the r+ on purpose; I want to see the updated patch with cloneInto...

Comment 8

[Bobby Holley (:bholley)]

(In reply to Boris Zbarsky [:bz] from comment #7)
> Comment on attachment 8728685 [details]
> MozReview Request: Bug 1213056 - change getStats to be maplike.
> 
> https://reviewboard.mozilla.org/r/39017/#review35789
> 
> ::: dom/media/PeerConnection.js:309
> (Diff revision 1)
> > +        get: function() {
> 
> OK, I checked with bholley.  We do not, in fact, want to expose the chrome
> function directly.  Use Cu.cloneInto with cloneFunctions=true to clone
> things involving accessors into the content compartment.

This isn't quite correct - cloneInto won't work with accessors AFAIK, because the structured clone algorithm invokes accessors while doing the clone. You'll need to use exportFunction for the accessors, but I also suggested using cloneInto(, { object literal}) to replace the createObjectIn and manual property definition.

Comment 9

[Byron Campen [:bwc]]

Comment on attachment 8728686 [details]
MozReview Request: Bug 1213056 - update tests to use maplike getStats.

https://reviewboard.mozilla.org/r/39019/#review35881

Comment 10

[Jan-Ivar Bruaroey [:jib] (needinfo? me)]

Comment on attachment 8728685 [details]
MozReview Request: Bug 1213056 - change getStats to be maplike.

Review request updated; see interdiff: https://reviewboard.mozilla.org/r/39017/diff/1-2/

Comment 11

[Jan-Ivar Bruaroey [:jib] (needinfo? me)]

Comment on attachment 8728686 [details]
MozReview Request: Bug 1213056 - update tests to use maplike getStats.

Review request updated; see interdiff: https://reviewboard.mozilla.org/r/39019/diff/1-2/

Comment 12

[Jan-Ivar Bruaroey [:jib] (needinfo? me)]

Got distracted with other things. Sorry for the delay, and thanks again for the pointers.

Comment 13

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 8728685 [details]
MozReview Request: Bug 1213056 - change getStats to be maplike.

https://reviewboard.mozilla.org/r/39017/#review41805

Comment 14

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/1c12251da027
https://hg.mozilla.org/integration/mozilla-inbound/rev/087f10d869e6

Comment 15

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/1c12251da027
https://hg.mozilla.org/mozilla-central/rev/087f10d869e6

Comment 16

[Kohei Yoshino [:kohei]]

Posted the site compatibility doc: https://www.fxsitecompat.com/docs/2016/rtcstatsreport-has-become-map-like-object/

Comment 17

[Jan-Ivar Bruaroey [:jib] (needinfo? me)]

Thanks! Note that there's a deprecation warning in console that triggers on legacy access (once per pc).