Bug 118844 – Make atob() / btoa() argument non optional

• Component: WebCore Misc.
• Status: RESOLVED FIXED

[Chris Dumez]

According to the latest specification, the argument to atob() / btoa()
should not be optional:
http://www.whatwg.org/specs/web-apps/current-work/multipage/webappapis.html#windowbase64

This patch makes WebKit behave according to the specification. The argument
is also mandatory in Firefox, IE10 and Blink.

atob() / btoa() are also moved to their own WindowBase64 interface which
the Window interface implements. This does not change the behavior but
this is closer to the specification and it will make exposing those
methods to workers easier later on.

Corresponding Blink revision:
https://src.chromium.org/viewvc/blink?revision=154419&view=revision

[Chris Dumez]

Created attachment 206981 [details]
Patch

[Kentaro Hara]

Comment on attachment 206981 [details]
Patch

LGTM. Thanks for keeping cross-browser compatibility.

[Alexey Proskuryakov]

Weird, why was the argument optional? Did you check svn history?

[Chris Dumez]

(In reply to comment #3)
> Weird, why was the argument optional? Did you check svn history?

From what I understand, a [LegacyDefaultOptionalArguments] extended attribute was added to the DOMWindow interface in <http://trac.webkit.org/changeset/89148>. This basically means that arguments are optional for all operations of the interface (Apparently used to be the default behavior in WebIDL.

Then, the interface-level extended attribute was replaced by an argument-level [Optional] extended attribute in <http://trac.webkit.org/changeset/92313> (no behavior change).

This optional extended attribute was then never removed.

Summary:
So this seems like a legacy from Web IDL, at a time when we did not throw when arguments were missing. These methods were simply never updated after that to match the specification.

With that in mind, any objection about me landing this change?

[Alexey Proskuryakov]

Comment on attachment 206981 [details]
Patch

Seems OK to me. Even though we're making exceptions happen where they didn't use to, it seems very unlikely that anyone calls these functions without arguments.

Changing behavior for null seems slightly more likely to affect sites, but also probably not bad enough.

[WebKit Commit Bot]

Comment on attachment 206981 [details]
Patch

Clearing flags on attachment: 206981

Committed r152859: <http://trac.webkit.org/changeset/152859>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Alexey Shvayka]

*** Bug 68618 has been marked as a duplicate of this bug. ***