kicad-developers team mailing list archive

[Cirilo Bernardo <cirilo_bernardo@xxxxxxxxx>]

----- Original Message -----

> From: jp charras <jp.charras@xxxxxxxxxx>
> To: Cirilo Bernardo <cirilo_bernardo@xxxxxxxxx>; "kicad-developers@xxxxxxxxxxxxxxxxxxx" <kicad-developers@xxxxxxxxxxxxxxxxxxx>
> Cc: 
> Sent: Friday, January 3, 2014 6:53 PM
> Subject: Re: [Kicad-developers] various patches
> 
> Le 02/01/2014 22:33, Cirilo Bernardo a écrit :
>>  ----- Original Message -----
>> 
>>>  From: jp charras <jp.charras@xxxxxxxxxx>
>>>  To: kicad-developers@xxxxxxxxxxxxxxxxxxx
>>>  Cc: 
>>>  Sent: Thursday, January 2, 2014 8:29 PM
>>>  Subject: Re: [Kicad-developers] various patches
>>> 
>>>  Le 02/01/2014 04:27, Cirilo Bernardo a écrit :
>>>>   Hi folks,
>>>> 
>>>>    I have put two sets of patches onto github for anyone to review 
> or apply.
>>>> 
>>>>   The (bzr) patches:
>>>>     export_idf3.patch :  Adds basic IDF3 export (board and cutouts / 
> holes 
>>>  only)
>>>> 
>>>>     export_vrml.patch : Improved VRML export; includes more 
> realistic 
>>>  coloring scheme, zone fills, improved eye candy. This patch is better 
> than the 
>>>  patch I previously posted here; this patch fixes a number of bugs, 
> eliminates a 
>>>  redundant class, and provides features such as the zone fills and 
> silver pads.
>>>> 
>>>>   URL: https://github.com/cbernardo/kicad-patches
>>>> 
>>>>   The patches are against Rev 4598.
>>>> 
>>>>   I would appreciate any comments on the patches.  Personally I 
> would like to 
>>>  see the patches incorporated with the bzr source. The vrml patch 
> includes a 
>>>  major rewrite of export_vrml.cpp and I'm confident users will find 
> the 
>>>  results to their liking.
>>>> 
>>>>   - Cirilo
>>> 
>>>  I committed the IDF patch
>>> 
>>>  vrml patch needs to be tested.
>>> 
>>>  Thanks.
>>> 
>>>  -- 
>>>  Jean-Pierre CHARRAS
>>> 
>> 
>> 
>>  Thanks Jean-Pierre,
>> 
>>   What kind of testing is required for the VRML patch? I'm happy to do 
> whatever I can.
>> 
>>   I tested the VRML export on 3 boards which I designed, some test boards to 
> ensure certain features are tested such as oval pad at angles different from 0, 
> 90, and on all the boards in the demo directory. Almost everything is OK except 
> for a few erratic issues with rendering some features, but so far I can find no 
> big problems. The erratic features I noticed so far include:
>> 
>>  1. On the pic_programmer demo board, some of the resistor silkscreen 
> outlines have a cone at the end rather than being flat.
>> 
>>  2. On one of my own boards, the silk outline of a connector has a line 
> forming a triangle rather than a rectangle; all other of the same connectors on 
> the board have the outline rendered correctly. I suspect this is related to (1)
>> 
>>  3. On the video demo board some of the thru holes and pads are rendered 
> strangely - the track extends two points into the pad and the pad has a 
> 'V' shape cutout on either side of the track.
>> 
>>  I have been looking into these issues and will spend more time trying to 
> create a simple board which replicates the problems. In general the erratic 
> features are not very obvious and will not harm the appearance. The resulting 
> files load into the whitedune viewer and Blender.
>> 
>>  - Cirilo
>> 
>> 
> 
> In fact it needs 2 changes in vrml_board to be compiled on mingw (I
> exactly use TDM-GCC-32, with differs slightly from mingw 4.7, and uses
> gcc 4.8.1)
> 
> 1 - in void SetGLError( GLenum errno ), errno collides a define in some
> mingw header. So change all occurrences of errno to something like error_id
> 
> 2 - the callback functions like:
> void vrml_tess_begin( GLenum cmd, void* user_data )
> are incorrectly defined.
> 
> they must use the CALLBACK type modifier
> (see for instance 3d_basic_functions.cpp):
> static void CALLBACK vrml_tess_begin( GLenum cmd, void* user_data )
> 
> I included fctsys.h to define CALLBACK (defined in a mingw include file)
> #include <fctsys.h>
> 
> I changed them to:
> 
> #ifndef CALLBACK
> #define CALLBACK // On linux, CALLBACK is not used, and not defined
> #endif
> 
> #define GLCALLBACK(x) (( void (CALLBACK*)() )&(x))
> 
> static void CALLBACK vrml_tess_begin( GLenum cmd, void* user_data )
> {
>     VRML_LAYER* lp = (VRML_LAYER*) user_data;
> 
>     lp->glStart( cmd );
> }
> 
> and later for each callback function:
> // set up the tesselator callbacks
> gluTessCallback( tess, GLU_TESS_BEGIN_DATA, GLCALLBACK(vrml_tess_begin) );
> 
> 
> After that, all works fine for me.
> 
> I suggest you to add holes defined by the board outlines.
> If you are using:
> BOARD::GetBoardPolygonOutlines() returns the external outline polygon,
> and a separate list of holes (polygons)
> (see 3d_draw.cpp)
> 
> 



Thanks for all your comments Jean-Pierre.

I have made the changes to vrml_board as requested.

The board cutouts were meant to be rendered but I overlooked it while testing and did not realize there were bugs. I fixed the bugs and the cutouts now render. I also fixed another bug (PCB text not rendered) I found while checking the VRML output against what I expected from my test board.

The latest patch has been pushed to github and is against Rev. 4601  I could not get any later revision to build (latest attempted: 4608).

I will try to set up MinGW as well so I can test in that environment as well.

Cheers,
Cirilo