Re: [E-devel] ecore_animator_pos_map broken with ACCELERATE

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

On Sun, May 6, 2012 at 10:34 PM, Michael Blumenkrantz
<mic...@gm...> wrote:
> 0.999987 consistently returns a value greater than 1.0 when using
> ECORE_POS_MAP_ACCELERATE.
>
> On Sun, May 6, 2012 at 1:11 PM, Carsten Haitzler <ra...@ra...>wrote:
>> On Sun, 6 May 2012 12:50:27 +0000 Michael Blumenkrantz
>> <mic...@gm...> said:
>>
>> ummm... i suspect a bug in your code. that accelerate mode has been in
>> edje for
>> years - ecore just copied the algorithm/code out into a shared func in
>> edje.
>> i've had test code for it - here is the dump of values "near 1":
>>
>> 0.909 -> 0.857
>> 0.919 -> 0.874
>> 0.929 -> 0.889
>> 0.939 -> 0.904
>> 0.949 -> 0.922
>> 0.960 -> 0.936
>> 0.970 -> 0.951
>> 0.980 -> 0.969
>> 0.990 -> 0.984
>> 1.000 -> 0.999
>>
>> never goes above 1.

There was a rouding and overflow issue in eina fixed point
implemantation of sin/cos that when getting close to 1 and above
0.999, it could wrongly round the value and think it was above 1.0. It
is now fixed, I just don't really get why no one noticed that before
as we use that in edje intensively.
-- 
Cedric BAIL