[Gzz-commits] journals tjl

[Tuomas J. Lukka]

CVSROOT:        /cvsroot/fenfire
Module name:    journals
Changes by:     Tuomas J. Lukka <address@hidden>        03/05/07 10:39:44

Modified files:
        .              : tjl 

Log message:
        more

CVSWeb URLs:
http://savannah.gnu.org/cgi-bin/viewcvs/fenfire/journals/tjl.diff?tr1=1.62&tr2=1.63&r1=text&r2=text

Patches:
Index: journals/tjl
diff -u journals/tjl:1.62 journals/tjl:1.63
--- journals/tjl:1.62   Tue May  6 09:25:50 2003
+++ journals/tjl        Wed May  7 10:39:43 2003
@@ -34,10 +34,13 @@
   do post-effects -- need to expose rectangle - mosaic.
     - what about laziness?
 
-- Screen-space dicing paperquad polishing
-    - homogeneous coordinates throughout transformation hierarchy!
-       - needed for improving texture coordinate quality
-       - Transform::isWGood()?
+- homogeneous coordinates throughout transformation hierarchy!
+    - needed for improving texture coordinate interpolation quality
+      in nonlinear views
+    - Transform::isWGood()?
+
+- GL vs non-GL split of coordsys; do the systems under the fisheye
+  using matrices
 
 Implementation
 """"""""""""""
@@ -50,6 +53,8 @@
 
 - canvas view 2d: show *where* (what dir.) there is more text / stuff
     - [mudyc] what's this?
+       - some visual indicator of where the rest of the stuff is on a canvas, 
i.e. if 
+         far right from anything I should see it. Something like a "radar"
 
 - libpaper: optimized paper generation in pbuffer!
 
@@ -133,30 +138,57 @@
 2003-05-10:
     - "DEADLINE": YLEM journal fillets
 
+2003-05-08:
+    - document dicing code, clean up the messy parts not
+      needed any more.
+
 2003-05-07:
     - [mudyc] how we can see from mouse click what 
       node(notes on pp canvases) was clicked if any?
         - put some example code to AbstractMainNode2D.doMouse
+           - MATCHER HIERARCHY!!!
 
     - [benja] buoyoing with two-focus interface
       (one main buoy manager, one buoy manager
       at the bottom of the screen showing pp papers)
 
-2003-05-06:
 
 =============
 Past:
 =============
 
+2003-05-07: Jkl
+    - Finish new dicing for paperquads - much better
+       - the important ideas for solving the problem:
+           1. don't try pure recursion. Splitting a triangle
+              must also split its neighbour. Having the real
+              structure of the mesh and an edge split operation
+              there is good
+           2. Don't forget source space. When splitting a triangle,
+              choosing the longest edge in *source* space produces
+              balanced meshes - quite often almost *structured* ones,
+              without any of the bookkeeping hassles. Avoids
+              lengthened triangles.
+           3. Use discrepancy and probe points liberally. When choosing
+              whether to split a triangle, the previous point allows
+              us now to just generate a new point at the center
+              and see how far it is from linear. The distance is then
+              the splitting criterion.
+              
+
 2003-05-06: [ tired + depressed; none of AGPU proposals taken ]
+    - nothing. Rested
+    - second place in monthly Blitz Go tournament of Helsinki Go Club
+       - feeling somewhat better
 
-2003-05-05: 
+2003-05-05: (LOTS)
     - new lego invoice to Virva
 
     - DEADLINE: IV'03 camera-ready copy!
          - formatting problems
 
     - DEADLINE: IV'03 controllers
+       - finished 01:45 
 
 
 2003-05-04: