[Bf-committers] DVCS again
Hello I've been following the discussion about switching to a distributed source control mechanism, and it's problem associated with binary data. According to previous chat here on the list, it is not yet decided which way we should go. What is the problem with binary data? Is it the big file size or slowness when checking for diffs? If big files are the problem, I just found this http://mercurial.selenic.com/wiki/LargefilesExtension which could help solve the problem. -- Micael Dias ___ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers
[Bf-committers] Interesting Article
An older article, but very interesting : http://www.vidimce.org/publications/lpics/ (Video and paper ) ___ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers
Re: [Bf-committers] Blender tangent space calculation
Hi! I think that I've finished this. Don't forget to look in do_multires_bake() in object_bake.c It shows you how to add the tangent layer: Yes, I've figured out how to create DerivedMesh object by Mesh object. And then I've created the tangent layer in this way. // get pointer to the already generated tangents pvtangent= DM_get_face_data_layer(dm, CD_TANGENT); But if mesh don't have a tex-coords, 'pvtangent' anyway is NULL For this case I weld normals only. I tried to export a mesh with tangents and normals into my app and seems that normal map looking correct. const int iGetNrVerts= data-mface[face_num].v4!=0 ? 4 : 3; On Thu, Nov 17, 2011 at 8:33 PM, Morten Mikkelsen mikkels...@gmail.comwrote: And flush_pixel() shows you how to traverse the buffer. Essentially the way it works in blender is there's always 4 of them per face whether it's a triangle or a quad. On Thu, Nov 17, 2011 at 9:31 AM, Morten Mikkelsen mikkels...@gmail.com wrote: Don't forget to look in do_multires_bake() in object_bake.c It shows you how to add the tangent layer: float *pvtangent= NULL; // create tangent vectors if not already created if(CustomData_get_layer_index(dm-faceData, CD_TANGENT) == -1) DM_add_tangent_layer(dm); // get pointer to the already generated tangents pvtangent= DM_get_face_data_layer(dm, CD_TANGENT); On Thu, Nov 17, 2011 at 8:49 AM, Eugene Minov minov@gmail.com wrote: The proper way to get the tangent layer can be seen in: https://svn.blender.org/svnroot/bf-blender/trunk/blender/source/blender/editors/object/object_bake.c Okay, I've found a 'multiresbake_get_normal' function for correct normals calculation. I also looked into 'DM_add_tangent_layer' function in: https://svn.blender.org/svnroot/bf-blender/trunk/blender/source/blender/blenkernel/intern/DerivedMesh.c According to it, the tangent layer is filling by tangents when created. So I do need only to access and welding them and normals? That's good if so. Today I had time to try understand how RNA's works. And I almost create and test python interface with collections for faces and indexed vertices with normals and tangents in it. Hopefully soon I'll start welding. If you need a free ultra simple welder there's one here -- http://jbit.net/~sparky/academic/welder/ You specify how many floats you have per vertex and it will weld for you. Okay, good one, I think I'll use it :) Thanks! On Wed, Nov 16, 2011 at 8:43 PM, Morten Mikkelsen mikkels...@gmail.com wrote: Sorry for confusing you here but I think I found a better reference for you since you'll be needing the tangents too and you are not supposed to be building them yourself. The proper way to get the tangent layer can be seen in: https://svn.blender.org/svnroot/bf-blender/trunk/blender/source/blender/editors/object/object_bake.c do_multires_bake() These are being fetched after the line: float *pvtangent= NULL; These are then read in the function flush_pixel() including the normal which is fetched from there using multiresbake_get_normal() which as you can see looks a lot like the GetNormal() function I pointed you to in DerivedMesh.c. Anyway, definitely use this file as your reference. I should have shown you this one from the beginning. If you need a free ultra simple welder there's one here -- http://jbit.net/~sparky/academic/welder/ You specify how many floats you have per vertex and it will weld for you. On Wed, Nov 16, 2011 at 5:32 AM, Eugene Minov minov@gmail.com wrote: If you can get hold of the dm (DerivedMesh) on the c side of things then I can show you how to get the correct normals and tangents and even help you get them welded should you want this. Ok! Sounds good to me :) So right now I in progress of checkout latest svn sources and compile blender. (had problems with net) Then first of I'll try to create test version of python/C interface. I've not decided yet what names I'll use for it. And finally will be trying to implement it looking in DerivedMesh.c, I think that it's realy are a good example. If or when I have a problem, I'll be glad to use your help :) Many thanks for your kind cooperation! On Tue, Nov 15, 2011 at 10:41 PM, Morten Mikkelsen mikkels...@gmail.com wrote: I don't know anything about Python but if you can get hold of the dm (DerivedMesh) on the c side of things then I can show you how to get the correct normals and tangents and even help you get them welded should you want this. On Tue, Nov 15, 2011 at 10:43 AM, Eugene Minov minov@gmail.com wrote: Yes, I absolutely agree, hard faces obviously must be
Re: [Bf-committers] Blender tangent space calculation
Sorry for extra line :) I wanted to ask about this line: const int iGetNrVerts= data-mface[face_num].v4!=0 ? 4 : 3; I think 'iGetNrVerts' may be incorrect with some cases when there is four vertices per face but the last one is points to first indexed vertex. Or there is no way that last index points to zero? On Fri, Nov 18, 2011 at 9:48 PM, Eugene Minov minov@gmail.com wrote: Hi! I think that I've finished this. Don't forget to look in do_multires_bake() in object_bake.c It shows you how to add the tangent layer: Yes, I've figured out how to create DerivedMesh object by Mesh object. And then I've created the tangent layer in this way. // get pointer to the already generated tangents pvtangent= DM_get_face_data_layer(dm, CD_TANGENT); But if mesh don't have a tex-coords, 'pvtangent' anyway is NULL For this case I weld normals only. I tried to export a mesh with tangents and normals into my app and seems that normal map looking correct. const int iGetNrVerts= data-mface[face_num].v4!=0 ? 4 : 3; On Thu, Nov 17, 2011 at 8:33 PM, Morten Mikkelsen mikkels...@gmail.comwrote: And flush_pixel() shows you how to traverse the buffer. Essentially the way it works in blender is there's always 4 of them per face whether it's a triangle or a quad. On Thu, Nov 17, 2011 at 9:31 AM, Morten Mikkelsen mikkels...@gmail.com wrote: Don't forget to look in do_multires_bake() in object_bake.c It shows you how to add the tangent layer: float *pvtangent= NULL; // create tangent vectors if not already created if(CustomData_get_layer_index(dm-faceData, CD_TANGENT) == -1) DM_add_tangent_layer(dm); // get pointer to the already generated tangents pvtangent= DM_get_face_data_layer(dm, CD_TANGENT); On Thu, Nov 17, 2011 at 8:49 AM, Eugene Minov minov@gmail.com wrote: The proper way to get the tangent layer can be seen in: https://svn.blender.org/svnroot/bf-blender/trunk/blender/source/blender/editors/object/object_bake.c Okay, I've found a 'multiresbake_get_normal' function for correct normals calculation. I also looked into 'DM_add_tangent_layer' function in: https://svn.blender.org/svnroot/bf-blender/trunk/blender/source/blender/blenkernel/intern/DerivedMesh.c According to it, the tangent layer is filling by tangents when created. So I do need only to access and welding them and normals? That's good if so. Today I had time to try understand how RNA's works. And I almost create and test python interface with collections for faces and indexed vertices with normals and tangents in it. Hopefully soon I'll start welding. If you need a free ultra simple welder there's one here -- http://jbit.net/~sparky/academic/welder/ You specify how many floats you have per vertex and it will weld for you. Okay, good one, I think I'll use it :) Thanks! On Wed, Nov 16, 2011 at 8:43 PM, Morten Mikkelsen mikkels...@gmail.com wrote: Sorry for confusing you here but I think I found a better reference for you since you'll be needing the tangents too and you are not supposed to be building them yourself. The proper way to get the tangent layer can be seen in: https://svn.blender.org/svnroot/bf-blender/trunk/blender/source/blender/editors/object/object_bake.c do_multires_bake() These are being fetched after the line: float *pvtangent= NULL; These are then read in the function flush_pixel() including the normal which is fetched from there using multiresbake_get_normal() which as you can see looks a lot like the GetNormal() function I pointed you to in DerivedMesh.c. Anyway, definitely use this file as your reference. I should have shown you this one from the beginning. If you need a free ultra simple welder there's one here -- http://jbit.net/~sparky/academic/welder/ You specify how many floats you have per vertex and it will weld for you. On Wed, Nov 16, 2011 at 5:32 AM, Eugene Minov minov@gmail.com wrote: If you can get hold of the dm (DerivedMesh) on the c side of things then I can show you how to get the correct normals and tangents and even help you get them welded should you want this. Ok! Sounds good to me :) So right now I in progress of checkout latest svn sources and compile blender. (had problems with net) Then first of I'll try to create test version of python/C interface. I've not decided yet what names I'll use for it. And finally will be trying to implement it looking in DerivedMesh.c, I think that it's realy are a good example. If or when I have a problem, I'll be glad to use your help :) Many thanks for your kind cooperation! On Tue, Nov 15, 2011 at 10:41 PM, Morten Mikkelsen mikkels...@gmail.com wrote: I don't know
Re: [Bf-committers] Cycles movement blur
It would be nice if this was built-in to how cycles worked, rather than a hack that you have to do manually. And even then, it's more a stop-gap solution. Ideally we'll (eventually) want randomly sampled motion blur, to eliminate strobing and to let you see the full motion blur (even if noisy) early in the render, so that you can tweak it in the viewport without having to do the full render (similar to DoF right now). Of course, I don't expect that terribly soon. ;-) Motion blur, particularly deformation motion blur, is a non-trivial problem in raytracing, if I understand it correctly. --Nathan On Tue, Nov 15, 2011 at 2:52 PM, Carsten Wartmann c...@blenderbuch.de wrote: It is already possible to get this kind of motion blur in cycles. Switch to blender internal and use the sampled motion blur panel to setup. Then back in cycles you get mblur samples. Combine with an animation of the seed value and lower the cycles samples by factor of mblur samples. As result you get mblur with nearly no more extended render time. Carsten Knapp magick.c...@gmail.com schrieb: I was just thinking today that Cycles could do motion blur without taking a big CPU hit. I don't know much about all this so I bet my ideas is wrong or old but I thought I would post it just in case. When we do animation the computer calculates tweens and moves stuff for us. It does this once per frame. Cycles does many render cycles per frame. It struck me today that if at each cycle pass you move the object you would get a blur effect. I was also thinking that you might move it every 100 out of a 1000 pass to get interesting stutter effects. I look forward to hearing if my idea was good or only trash. BTW I would love to see a pause on the main render. Cycles eat my CPU alive and does not let me do much in the background. A nice/slow button might be good too but I am sure I am not the first to say this. Thanks for all the really cool work!! I love cycles, dynamic paint, motion tracking and ocean sim! All the other stuff is more in the background but I am sure I will be loving it too once I find it! -- Douglas E Knapp Creative Commons Film Group, Helping people make open source movies with open source software! http://douglas.bespin.org/CommonsFilmGroup/phpBB3/index.php Massage in Gelsenkirchen-Buer: http://douglas.bespin.org/tcm/ztab1.htm Please link to me and trade links with me! Open Source Sci-Fi mmoRPG Game project. http://sf-journey-creations.wikispot.org/Front_Page http://code.google.com/p/perspectiveproject/ ___ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers -- Mühsam auf dem Telefon getippt :-) ___ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers ___ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers
Re: [Bf-committers] Blender tangent space calculation
I wanted to ask about this line: const int iGetNrVerts= data-mface[face_num].v4!=0 ? 4 : 3; Or there is no way that last index points to zero? Correct, this is how blender does it everywhere in its code-base. Super glad to hear you're getting useful results! Be sure to submit a patch for review. ___ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers
Re: [Bf-committers] DVCS again
There is something similar for git too: http://git-annex.branchable.com/ Original-Nachricht Datum: Fri, 18 Nov 2011 17:37:40 + Von: Micael kam1k...@gmail.com An: bf-blender developers bf-committers@blender.org Betreff: [Bf-committers] DVCS again Hello I've been following the discussion about switching to a distributed source control mechanism, and it's problem associated with binary data. According to previous chat here on the list, it is not yet decided which way we should go. What is the problem with binary data? Is it the big file size or slowness when checking for diffs? If big files are the problem, I just found this http://mercurial.selenic.com/wiki/LargefilesExtension which could help solve the problem. -- Micael Dias ___ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers -- Dr. Lars Krueger Empfehlen Sie GMX DSL Ihren Freunden und Bekannten und wir belohnen Sie mit bis zu 50,- Euro! https://freundschaftswerbung.gmx.de ___ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers
Re: [Bf-committers] Please help me debugging
Oh dear, well, I used 2.6.5 for running the script - didn't know that that would make such a difference. With python 3.2 it works just fine - finally! Thank you for your help Campbell ! Rainer ___ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers
Re: [Bf-committers] Please help me debugging
Python3 is default on my system so I forget about this sometimes. Glad to hear its working (though there _was_ a real bug with utf8 paths I found because of your mail). Added checks for python3 r41980 so this doesn't happen to anyone else. On Sat, Nov 19, 2011 at 10:51 AM, Rainer Hohne raho...@googlemail.com wrote: Oh dear, well, I used 2.6.5 for running the script - didn't know that that would make such a difference. With python 3.2 it works just fine - finally! Thank you for your help Campbell ! Rainer ___ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers -- - Campbell ___ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers
Re: [Bf-committers] UTF8 text input status (on OS X)
Hi Brecht, Tested the patch here it it works great. To test it from any OSX: option+e+a = á option+i+o = ô The only downside is that does not address the problem of korean, japanese, ... probably because of the problems presented by Alex. It's already an improvement though. -- Dalai (a related OSX issue I have here: hold space and press backspace in the Python Console. It will produce a lot of the 'square' chars). 2011/11/1 Campbell Barton ideasma...@gmail.com: On Wed, Nov 2, 2011 at 1:08 PM, Alexandr Kuznetsov kuzsa...@gmail.com wrote: Hi c) Typing with dead keys (e.g. ¨ + o = ö) didn't work for me. I managed to get it working using NSTextInput, patch here if people want to test, I'm not very confident that this doesn't break something, so some testing would be great: http://www.pasteall.org/25983/diff I think this problem is much bigger here. The dead keys are typed before a real key, but the composed UTF chars are store in the reverse order. Even if we have a buffer for that, it won't solve many problems. For example Chines and Japanese inputs require more closer interaction with the text, rather than simply typing character. Different input methods exist for different languages which are natively supported by OS. But because Blender doesn't use native text fields, they must be implemented by ourselves. For example, MS Pinpyin replaces typed latin characters into chinese when the whole word is complete. Therefore Blender must expose more text interface to Ghost. Ghost must be able to get the current line, cursor, visual position of editing text and then push edited text back. I probably can take care of this project, but it won't be until end of December. Plus I don't know any hieroglyphics languages. There is multiple cross platform implementations of IME. When we were discussing this with JesterKing, he proposed to use SDL. But although it might have ready solution, it is probably hard to integrate in our Ghost, as all SDL must have its own loop. (Anyway, when IME will be implemented, we might get speech recognition and handwriting for free!!) If you want to test SDL text input you could try WITH_GHOST_SDL to build ghost against SDL rather then x11/win/cocoa. SDL1.3 has a Text Input event, but the text editing state needs to be set first so you're right that ghost would need to be aware and have some equivalent to SDL_StartTextInput / SDL_StopTextInput, rather then just passing events like it does now. a) Uppercases character umlaut (e.g. Ü) are not rendered correctly in the user interface, they do show in 3D text and the text editor. Appears to be a text rendering problem that would happen on all platforms? Yes. It is a problem on Windows also. Moreover it is a problem exist with precomposed characters. Maybe it is problem of clipping? Anyway, maybe it is good time to switch char render library as was discussed earlier to support bidirectional text. Maybe there are more latent problem with the rendering of hard characters like Chinese or multiple compositions. Best Regards, Alex -- - Campbell ___ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers ___ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers