[Bf-committers] Armature total mess?

[bjornmose]

Chris Want wrote:
> Ton Roosendaal wrote:
> 
>> Current conclusion; it not only has to be rebuilt from scratch, but 
>> we  also have to drop any attempts for compatibility... and try to get 
>> a  conversion coded *after* a new design has proven to work. If 
>> that's  even possible.
>>
>> For the devs here who ever tried to grasp it (Hos, Theeth); is this  
>> conclusion correct? Or should I not give up and try to reveil the  
>> hidden treasures of the current implementation?
> 
> 
> Hi Ton,
> 
> Welcome to my hell!
> 
> I think it is pretty confusing too. I spent a lot
> of time trying to optimize the speed of the darn
> thing, and after doing that I still felt that I
> didn't understand it very well (mind you, I find a
> lot of the blender code to be confusing).
> 
> I would say start fresh if you feel the current code is
> too hard to extend... just please don't bring back IKA's!
> While you're at it, I think it would help to re-evaluate
> the subsurf system so that it supports partial updating
> of the mesh during deformation (i.e., currently, when you
> manipulate a character's pinky, the whole character's subsurf
> has to be recalculated on every redraw, not just the pinky).
> This subsurf recalculation was the bottleneck after
> my speed optimizations (my optimizations involved
> stopping blender from recalculating the entire armature's
> pose when you just manipulate the pinky). Maybe this is
> something for a 'displist refactor'. Anyways, there is no
> doubt: an armature overhaul is a *huge* job that touches
> a lot of blender code and will take a lot of time (hehe, let
> me know if you want tuhopuu commit rights >=P).
 From softbodies project experiance and a general confusement on what is 
  worthy to be in BF-tree versus what is to be considered 'work in 
progress' I'd like to have the 'armatures redesign', if reusing old code 
or not, in tuhopuu. I know it's a pain to have multiple states of code, 
but still i stick to the concept of having an 'offical stable relieable' 
  user fit version.

*** generic software development managers (me) dialog  other actors 
developer (d)
boss (boss)
---
d: can we relaese a new version?
me: why, i have no project tagged finished for now?
d: just a tiny little bug fix on feature XXX
me: other work done in code?
d: dunno what ABC and others did but it seems to work ..
me: did someone test if this does'nt break workflow for 99% of useres
d: hum .. no .. but is's a tiny litte bug fix .. so can we release ..
me: ... (scratches chin)
d: customer XYZ needs it badly ..
boss: customer XYZ won't pay the invoice unless he has the fix. We need it
me: can we give customer XYZ a special build?
boss: i need a build including the work do ABC and that new gimmick you 
did lately. otherways project on customer TZU is lost.
me: so developer ABC confirmed project for cusromer TZU is done?
boss: he says it works
....
and so on
....

me/
a. hopes there is no mayor damage in code
b. hopes there is enough beer in the fridge to drown all the issues i 
can think of ..
***

away the fun part, I'd prefer BF-tree to be the 'customers/users' one. 
Means: they can work with a BF-build without any nasty surprises at 
*any* time. If was strict only confirmed && tested bug fixes and 
finished && approved new features should go there.

Yup.. and sorry but, Ton this goes to you in person:
I would have felt much more comfortable if softbodies project went to 
tuhopuu first.

> 
> Out of curiousity: do you have any experience with the
> armature system from the user side of things? While the
> code is sure hell, I guess if there is a hidden treasure
Yeah .. come up with a new design .. and all the users ( including me ) 
will complain about all the things they/we've been able to do.
> it would be that it allows users to do a lot of cool stuff
> (although the system looks a bit dated compared to other
> modern animation systems).
> 
> Chris
> 
> P.S. before starting over, perhaps you could send Reevan
> a mail and ask a few questions?
> _______________________________________________
> Bf-committers mailing list
> Bf-committers at projects.blender.org
> http://projects.blender.org/mailman/listinfo/bf-committers
> 
> 

O.K. said above,
I'd vote for a rewrite of armature code
I can offer:
a. some (growing) knowledge of the animation system (code and user wise)
b. working on a IK solver system supporting 'multi ended / tree like 
IKs' .. current system *only* allows *2* ended IK chains.
c. keeping an eye on *true* artists needs versus *true* 'clean' code && 
data structs/dependancies  .. humm .. there was a little spike on IRC 
which made me write it like this :)

-Ole-