Re: [E-devel] E SVN: kakaroto trunk/edje/src/lib

[Cedric B. <ced...@fr...>]

On Mon, Jan 9, 2012 at 2:06 PM, Bruno Dilly <bd...@pr...> wrote:
> On Sun, Jan 8, 2012 at 2:32 PM, Cedric BAIL <ced...@fr...> wrote:
>> On Sun, Jan 8, 2012 at 3:47 PM, ChunEon Park <he...@na...> wrote:
>>> I think both are no problems if it has a documentation.
>>> But your patch may break applications already released.
>>> It will be better to apply your patch when major version is changed.
>>
>> As I say, current behaviour is undefined. If you go out of an
>> animation (defined in the edj itself) in any state (hidden, moved,
>> resized, whatever), it will stay in that state. But this is completly
>> random and not defined (as in, depend on an external file). Now I do
>> like the raster proposal with an orphaned flag as it is the only sane
>> way to detect any leak. Relying on an undefined visual artefact would
>> not help at all.
>
> It isn't documented. But it's defined, IMHO, since you can predict it.
> As you said, in an animation it will keep the state, if it was
> visible, it will stay visible.
> So applications can be considering a unswallowed object will be
> visible, since it was visible, and now it will be hidden.

No, as it is defined in the theme, it doesn't depend on the
application. If you change the theme, the animation, anything in the
.edj, it will change the behaviour in the application itself. It's
full of race condition. There is no sane way to expect any kind of
behaviour in the app. It is definitivly an undefined behaviour, as
their is no way you could know the state of the object without
requesting it after the unswallow.

> Despite it wasn't documented when a person tried to unswallow she
> realized it was visible, or maybe she looked at the code.
> Consequently, she didn't forced to show the object with
> evas_object_show(), what will be mandatory now. So Hermet is correct,
> it can break applications and won't be something simple to see like an
> API breakage. People will spend a good time to realize what changed.
> Sure, we can help with that warning as we can (changelog, msg to mail
> list, docs, etc).

This is a bug in the application, that rely on luck more than anything
if it worked at any point in time. As I said, if any application rely
on it now, just let me touch the edj and it will be broken completly.
So no, this is not a big change. And I really doubt that any
application outside as this kind of bug that goes unnoticed.
-- 
Cedric BAIL