plib/TODO_AFTER135

+============+
| TO BE DONE |
+============+

A tasks and bugs are now being tracked by sourceforge. Please 
goto: http://sourceforge.net/projects/plib/

  * Have a look into known_bugs and fix them.

  * Lines in ASCII files are terminated...

          With '\n' under UNIX/Linux
          With '\r' under MacOS
          With BOTH under Windoze.

    Hence, UNIX/Linux does no conversions of either '\r' or '\n'.
    and    MAC swaps '\r' and '\n' on input.
    and    Windoze dumps '\r' if it follows a '\n'.

    This is a mess - so I'm changing all the ASCII I/O code
    to allow either or both '\n' or '\r' and I'm reading the
    ASCII files in BINARY mode.
 
  * Array deletion requires '[]' after 'delete' on Mac.

  * Some of the MSVC-project files for plib_examples seem to be 
    broken, for example some miss a "winmm.lib".

  * Check whether the new ssgStripify works with TuxKart
    See Steves post from 11.12.2000 15:12 for a problem description



The abbreviation NIV14 means not in Version 1.4.






Bugs:
=====

- When creating fgfsTux, sometimes .ac, .dxf,
  .ase and .obj saved zero objects, although 
  there were objects. For example, PLIB 
  created this .ac file:
  ----------- snip ------------
  AC3Db
  OBJECT world
  kids 0
  OBJECT group
  kids 0
  ----------- snip ------------
  This seems to happen after ssgFlatten.
- WK: Create a sphere in ppe. Save it as .ASE. See that values like
  diffuse colour are cr*p. Try to load it. It crashes. This could be 
  one or - probably - two bugs. I don't think this is a ppe problem. 
  The loader complains that number of faces is -2.
  The problem seems to be the writer.

  *new remark*: About the diffuse values: I think it has to do with
  colour material, a loader (.ASE?) loading colours into a colour
    list and writing colours from the ssgSimpleState where they are
    undefined. I think the same bug is in .AC 

- WK: I have a crash in ssgFlatten if I load .ac files with 
  unused materials (not reproducable?)
- Loading and saving lines in .OBJ doesn't work. Maybe NIV14.
- Sam wrote:
  Has anyone here debugged with plib under Windoze 2000 and MSVC++?
  I get a whole bunch of
  "Free Heap block modifed after it was free"
  warnings with the ssgFlatten and ssgStripify.
  I'll try and hunt this one down.
  ...
  I think it may be because I'm using MFC which (if I remember 
  correctly) enables a bunch more memory checks.  Also I'm doing
  #ifdef _WIN32
      #ifdef _DEBUG
          #define new DEBUG_NEW
      #endif
  #endif
- Not all loaders use ssgLoaderOptions::begin. Don't all have to?

- Search for todo, fix, fixme, kludge.


  
Misc:
=====

- Look whether scaling works. [see current discussion]
  If not and if we can't/don't want to make it work, for example 
  because of performance issues we need a warning in the docs and 
  IMHO PLIB should "write out" a warning if an unallowed matrix is
  sent to it.
  Update: Scaling does not work (intentionally). Uniform scaling may be
  enabled by uncommenting "radius *= sgLengthVec3(m[0])" in
  sgSphere::orthoXform. It is currently disabled because that extra
  calculation would degrade performance on non-scaled matrices.
- Reduce lint warnings. Lint is an error checking tool that gives warnings 
  where things *could be* wrong as opposed to the compiler that tells 
  you where they are wrong.

- It would be great if we had one or even two working native .ssg 
  file formats. Currently (16.12.2001) the format work most of the time, but not all 
  entities are implemented.
- We should write the GetWrapU/GetWrapV function, since 
  loading /writing .ssg files can't work 100% without it.
- When saving, there is often a warning that "ref count doesn't 
  tally with parent count"
- Go through the mailing list for unfinished business.
- Someone, preferably a native english speaker with ssg-knowledge,
  should go through the new ssg-doc and fix any mistakes. It should be 
  fairly complete by now.
- Add Per's new Formats to the doc?
- For the other parts of plib, people should bring the doc up to 
  date.
- There is a bug in the 3Dfx driver for Linux (tdfx_dri-4.0.1-1) that
  causes the "complex" example program to crash in "fxSetupBlend".
  This is fixed in more recent DRI snapshots (requires kernel 2.4.x).


Probably after 1.4.0:
=====================
- In fgfsTux, not a large object, you get a DList overflow
  with the standard values for its size. Can we make its
  size dynamic?
- Look at handling of normals. For example, loaders, 
  stripify etc should only recalculate missing normals.
  Have *one* function that recalculates normals.



It would be nice if people would volunteer for tasks.