296821 - restore() should reset state to initial values

Status: RESOLVED INVALID

(Core :: Graphics: Canvas2D, enhancement)

Description

[ArronM (:paper)]

Before the Canvas specs were rewritten based on Gecko, it said for restore:

"If there is no saved state, the method resets the context's drawing state to
its initial values."

I think this is a valid, logical, and usefull action.  All it would require it
pushing the state at startup, and never popping it (pull it, but keep it in the
state stack)

Comment 1

[Hixie (not reading bugmail)]

The change was made based on Safari, not Gecko.

Comment 2

[Mike Shaver (:shaver emeritus)]

Resolving this (even if as WONTFIX) needs to happen before 1.1, says me.

Comment 3

[ArronM (:paper)]

(In reply to comment #1)
> The change was made based on Safari, not Gecko.

I stand corrected, however, I still think it's a good idea.  Just because Safari
lacked the insight to think of this feature, someone (I assume you, Hixie, since
the spec is in your name) did.

Resolving as wontfix?  Why? The fix isn't impossible and the the spec is not a
standard yet.  In fact, it changes very often (and, I might add, without a
revision trail).


This really is an extension of my arguement that a spec shouldn't be written
soley based on an implementation built behind closed doors.  I mean, why can't
the spec still say restore should reset the context's drawing state, even if the
current implementations don't support it yet?  That goes for any other good
ideas that were/are written into the spec but Safari currently does have. 
Sounds to me like the big 3 browsers just want to gloat "we support 100% of the
canvas tag! [..because we wrote the spec to exclude anything we hadn't coded]"

Comment 4

[Mike Shaver (:shaver emeritus)]

I'm not saying that it _should_ be WONTFIX, just that FIXED or WONTFIX need to
happen before 1.1 goes out.

What good is a specification that nobody implements?  You then need to know a)
what the spec says, and b) what you really have to be compatible with, if you
implement it.  (Or what subset of the spec you can really use, if you're trying
to write something that actually uses it.)

It's not about gloating, but WHATWG process and spec issues belong on the WHATWG
list, not in our bugzilla, so please limit comments here to technical discussion
of what Mozilla should do here.  If restore()-before-save()'s behaviour is
undefined, we're free to make it reset to default values; that doesn't require a
spec change, and we should discuss it here on its specific merits.

Comment 5

[ArronM (:paper)]

(In reply to comment #3)
> ideas that were/are written into the spec but Safari currently does have. 
s/does have/does not have/

The problem with a volitile spec is that it's just that.. volitile.  Yesterday
it said restore reverts to default, today it says no action, tommorow, who knows?

(In reply to comment #4)
> What good is a specification that nobody implements?  You then need to know 

heh, what good is implementing a spec when it keeps changing?  I think this is
why Gecko has shied away from evolving, incomplete specs in the past.

Comment 6

[Hixie (not reading bugmail)]

restore()-before-save()'s behaviour is not undefined, it's currently defined to
be a no-op, it used to be defined to be a reset. The spec can change, it's a
draft. It says so right at the top of the spec in its status section. In any
case, if you have comments on the spec itself, please send them to
[e-mail]; similarly if you have comments regarding Mozilla's process
please take them up with [e-mail] or some other appropriate forum.

I agree that reset would be useful here. Mozilla used to crash, we changed
Mozilla to do the same as Safari because that was easiest, and made sense, and
the spec changed to match. If Vlad thinks that the reset behaviour is better,
then he'll decide to change to that, and the spec can be updated, assuming Apple
also agree.

Comment 7

[Stuart Parmenter]

I disagree that this is generally useful.  If you want to get back to the
default state put a save() as the first line of your code and be done with it. 
You can then restore the that point and save again...

Comment 8

[Asa Dotzler [:asa]]

renominate for b5 if this is critical.

Comment 9

[Stuart Parmenter]

this should definatly not be blocking the branch. it should be marked invalid,
if anything.

Comment 10

[Takeshi Kurosawa :takenspc]

The spec has been changed.
http://www.whatwg.org/specs/web-apps/current-work/multipage/section-the-canvas.html#drawing0
http://www.w3.org/TR/2008/WD-html5-20080122/#drawing0

"If there is no saved state, the method must do nothing."

-> INVALID

related: bug 412530