605660 - Object.init() was recently hidden by API versioning, but existing content relies on it

Status: RESOLVED FIXED

(Tamarin Graveyard :: Virtual Machine, defect)

Description

[Steven Johnson]

In the sad-but-true department: At least one Facebook game ("Crazy Cow Music Game") has a call to "::init()" in one of its classes constructors. Prior to Argo, this happened to work, because it resolved (eventually) to Object.init(), which is (mostly) harmless to call repeatedly. However, in Argo, this call was hidden (via AIR_SYS) and so this game throws an RTE during startup.

Not sure if we *should* fix this, since the content was in the just-getting-luck category, but it introduces an incompatibility...

Comment 1

[Steven Johnson]

Correction: bug is not in Argo (change was introduced after Argo lockdown) but is in Spicy/Serrano.

Comment 2

[Edwin Smith]

It might be possible to contact the devs of that game proactively if this is a unique case.

Comment 3

[Steven Johnson]

I'll bet you a dollar we'll find other SWFs that do this. (The question is how many, of course.)

Comment 4

[Edwin Smith]

Is there a fingerprint we can scan for using Brightspot?

Comment 5

[Steven Johnson]

(In reply to comment #4)
> Is there a fingerprint we can scan for using Brightspot?

Maybe, but probably not reliably -- depends on the code being used to invoke "init" and how it happens to resolve (or not).

Comment 6

[Lars T Hansen]

Since init() has an effect - it restores the standard values of Object.prototype in the event they have been updated by the program - it could in principle be called deliberately by content to achieve just that effect.

Comment 7

[Steven Johnson]

(In reply to comment #6)
> Since init() has an effect - it restores the standard values of
> Object.prototype in the event they have been updated by the program - it could
> in principle be called deliberately by content to achieve just that effect.

True, but this seems even less likely than the existing, apparently-accidental use case.

Comment 8

[Steven Johnson]

For Spicy, I'm going to revert this change in the name of compatibility.

Comment 9

[Steven Johnson]

Attachment: Patch

Provide an empty stub with the same signature as the old Object.init.

Comment 10

[Lars T Hansen]

Comment on attachment 489284 [details] [diff] [review]
Patch

R+ with reservations as outlined in earlier comments - this fix does not actually revert us to the old behavior, but it's probably the right trade-off.

Comment 11

[Steven Johnson]

TR 5535:3b7ea1e01705