RTCIceCandidate constructor shim throwing error if ufrag given for Firefox beta (111) about adapter

[fippo]

Thank you for backing this out @jan-ivar! I agree with the plan to reland we lets still try to get #1123 out

[fippo]

@jan-ivar Firefox breaking like this doesn't sound good? Past attempts by Chromium to enforce read-onlyness of attributes (in particular RTCSessionDescription's sdp) turned out to break too much for no gain.

[fippo]

Related Firefox bug: https://bugzilla.mozilla.org/show_bug.cgi?id=1812346

[jan-ivar]

The bug is in adapter.js here:

  const nativeCandidate = new NativeRTCIceCandidate(args);
  const parsedCandidate = SDPUtils.parseCandidate(args.candidate);
→ const augmentedCandidate = Object.assign(nativeCandidate, parsedCandidate); // TypeError

TIL the difference between:

1. assignment to a read-only attribute, which is ignored and doesn't throw, vs.
2. Object.assign which does: "In case of an error, for example if a property is non-writable, a TypeError is raised".

This fiddle shows the latter throwing in all browsers including Firefox 111+, so unfortunately this shim, which was specific to Firefox, was relying on non-standard behavior of Firefox 110 and earlier.

We'll back out the fix that broke adapter in Firefox 111 beta for now, to give adapter time to fix this, and for all users of adapter to update (which may be a while).

[fippo]

The fix is going to say

const nativeCandidate = new NativeRTCIceCandidate(args);
const parsedCandidate = SDPUtils.parseCandidate(args.candidate);
const augmentedCandidate = Object.assign({}, nativeCandidate, parsedCandidate);

right?

Adoption is going to be a problem, there is still a lot of usage on version 7 and not even 8.
You might consider not doing this until you implement attribute parsing, in particular foundation which is what prevents this from happening in Chrome.

[jan-ivar]

That would return the wrong type I fear. [Deleted: pilot error since I forgot RTCIceCandidate has a custom toJSON]

Agreed about adoption.