merge

Merge remote branch 'upstream/visitor' into visitortests

Conflicts:
	src/export.cc
	src/openscad.cc
	src/polyset.cc
	src/transform.cc
	tests/CMakeLists.txt
	tests/FindGLEW.cmake
	tests/csgtermtest.cc

10 files changed, 21 insertions, 26 deletions

diff --git a/doc/OpenSCAD-classes.graffle b/doc/OpenSCAD-classes.graffle
index ef8c1b8..e067521 100644
--- a/doc/OpenSCAD-classes.graffle
+++ b/doc/OpenSCAD-classes.graffle
diff --git a/doc/OpenSCAD-classes.pdf b/doc/OpenSCAD-classes.pdf
index 5870db0..04b62a9 100644
--- a/doc/OpenSCAD-classes.pdf
+++ b/doc/OpenSCAD-classes.pdf
diff --git a/doc/OpenSCAD-compile.graffle b/doc/OpenSCAD-compile.graffle
index bb505c3..9a09fc3 100644
--- a/doc/OpenSCAD-compile.graffle
+++ b/doc/OpenSCAD-compile.graffle
diff --git a/doc/OpenSCAD-compile.pdf b/doc/OpenSCAD-compile.pdf
index 2c9730d..efc7134 100644
--- a/doc/OpenSCAD-compile.pdf
+++ b/doc/OpenSCAD-compile.pdf
diff --git a/doc/OpenSCAD-csg.graffle b/doc/OpenSCAD-csg.graffle
index ca0a0ff..507c694 100644
--- a/doc/OpenSCAD-csg.graffle
+++ b/doc/OpenSCAD-csg.graffle
diff --git a/doc/OpenSCAD-csg.pdf b/doc/OpenSCAD-csg.pdf
new file mode 100644
index 0000000..0304cf7
--- /dev/null
+++ b/doc/OpenSCAD-csg.pdf
diff --git a/doc/OpenSCAD-polygons.pdf b/doc/OpenSCAD-polygons.pdf
index d2eaa3b..ac18baf 100644
--- a/doc/OpenSCAD-polygons.pdf
+++ b/doc/OpenSCAD-polygons.pdf
diff --git a/doc/TODO.txt b/doc/TODO.txt
index dd9932c..97af30a 100644
--- a/doc/TODO.txt
+++ b/doc/TODO.txt
@@ -6,6 +6,7 @@ o Tesselation via GLU sometimes produces strange results
 o Export STL: Exports existing CGAL model even though the current model is changed, but not CGAL rendered
 o Look into the polygon winding and rotate_extrude() problem reported by Britton
 o CGAL Aff_transformation_3 doesn't support non-affine transformations (non-aff-matrix.scad)
+o 2D union of polygons with a touching vertex doesn't work. see testdata/scad/bugs/polygon-touch.scad
 
 STL Import BUGS
 ---------------
@@ -39,12 +40,6 @@ Using STL-imported models is tricky and triggers multiple issues:
 CRASH BUGS
 ----------
 o Broken polyhedron() entities are not correctly detected and cause CGAL segfaults
-o linear_extrude(50) square([5,0]);
-o union() {
-    linear_extrude(height=10, twist=90) circle(5);
-    translate([7,-5,0]) linear_extrude(height=10, twist=180) polygon(points = [[0,0], [10,0], [5,10]]);
-  }
-o non-convex minkowski example from chrysn fails with an exception
 
 USER INTERFACE
 --------------
@@ -78,9 +73,12 @@ o 3D View
   - overlay indicator displaying current view mode
   - OpenCSG rendering: Coincident surfaces causes z-buffer fighting. Is this somehow
     avoidable tuning the depth tests in OpenCSG? 
-  - Make the 10.000 element OpenCSG klimit configurable (Preferences) ?
+  - Make the 10.000 element OpenCSG limit configurable (Preferences) ?
   - Use OpenGL picking to facilitate ray-tracing like features like measuring
     thicknesses, distances, slot thicknesses etc.
+  - When specifying a transparency with the color() statement,
+    the object is not sorted and will be rendered wrongly
+  - Add option to change lights, e.g. add an optional camera light
 o Editor wishlist
   - More infrastructure for external editor (allow communication from the outside)
   - Preferences GUI for the features below
@@ -133,20 +131,16 @@ o 2D Subsystem
 o Built-in modules
   - extrude*: Allow the base 2D primitive to have a Z value
   - rotate_extrude(): Allow for specification of start/stop/sweep angle?
-  - Convex hull of multiple 2D or 2D objects
+  - Convex hull of 3D objects
 o Advanced Transformations
   - Add statement for refinement via surface subdivision
   - Add statement for intersections in cartesian product of childs
+  - non-convex minkowski example from chrysn fails with an exception (testdata/scad/bugs/minkowski-assert.scad)
 o Function-Module-Interface
   - Pass a module instanciation to a function (e.g. for a volume() function)
   - Pass a function to a module instanciation (e.g. for dynamic extrusion paths)
 o Language Frontend
-  - include statement doesn't support nesting. This can be fixed by
-    keeping a nested stack of current input files in the lexer. See
-    the "Flex & Bison" O'Reilly book, "Start States and Nested Input
-    Files", page 28, for an example.
   - Allow local variables and functions everywhere (not only on module level)
-  - allow any expression to be evaluated as boolean (!0 = true, 0 = false)
   - Rethink for vs. intersection_for vs. group. Should for loops
     generate child lists instead, and make these passable to other
     modules or accessible by child()?
@@ -163,8 +157,6 @@ o Mesh optimization on STL export
   - Remove super small triangles (all sides are short)
   - Replace super thin triangles (one h is short)
 o Misc
-  - When specifying a transparency with the color() statement,
-    the object is not sorted and will be rendered wrongly
   - Go through default values of parameters (e.g. cube() has x,y,z=1 while linear_extrude() has height=100)
   - Add support for symbolic names to child() statement 
   - Add 'lines' object type for non-solid 2d drawings
@@ -176,7 +168,6 @@ o Misc
 o Grammar
   - dim->name -> dim->label
   - A random(seed) function
-  - import_*() -> *_import() (consistent prefix vs. postfix)
   - linear_extrude()/rotate_extrude(): Cumbersome names? -> (extrude, revolve, lathe, sweep ?) 
 o Hollow donut problem
   When extruding a 2D CSG tree (e.g. a polygon with a hole), the hole
@@ -184,7 +175,8 @@ o Hollow donut problem
   extrusions, this has only a minor visual impact, but for rotate
   extrusion, the resulting CGAL models will lose the hole. The OpenCSG
   rendering keeps the hole, but renders slightly incorrect.
-
+o CGAL issues
+  - CGAL doesn't handle almost planar polygons. Consider splitting into triangles ourselves. See WillamAdams/dodec.scad
 
 
 IDEAS FOR LANGUAGE CHANGES
@@ -223,11 +215,6 @@ o Caching and MDI looks suspicious when the code relies on external resources
   which might be loaded from difference locations in different documents
   -> we might get a false cache hit
 o Are contructs like "child(0)" cached? Could this give false cache hits?
-o Write some cmd-line apps that dump an openscad file to misc. formats
-  (dump, stl, dxf)
-o Write a simple test script that collects verified and current STL renderings
-  and displays them side-by-side or smth.
-o Write simple driver scripts for comparing output of above command
 o Collect "all" available OpenSCAD scripts from the internets and run the integration
   tests on them all
 o dumptest tests:
diff --git a/doc/testing.txt b/doc/testing.txt
index 417a3b9..b64d5a4 100644
--- a/doc/testing.txt
+++ b/doc/testing.txt
@@ -16,12 +16,13 @@ Adding  a new regression test:
 
 1) create a test file at an appropriate location under testdata/
 2) if the test is non-obvious, create a human readable description of the test in the same directory (e.g testdata/scad/mytest.txt)
-3) if a new test app was written, this must be added to tests/CMakeLists
-4) run the test with the environment variable TEST_GENERATE=1, e.g.:
+3) if a new test app was written, this must be added to tests/CMakeLists.txt
+4) Add the tests to the test apps for which you want them to run (in tests/CMakeLists.txt)
+5) run the test with the environment variable TEST_GENERATE=1, e.g.:
    $ TEST_GENERATE=1 ctest -R mytest
    (this will generate a mytest-expected.txt file which is used for regression testing)
-5) manually verify that the output is correct (test-data/scad/mytest-expected.txt)
-6) run the test normally and verify that it passes:
+6) manually verify that the output is correct (tests/regression/<testapp>/mytest-expected.<suffix>)
+7) run the test normally and verify that it passes:
   $ ctest -R mytest
 
 Note that test files which don't have an *-expected.<suffix> file will
diff --git a/doc/visitor-changes.txt b/doc/visitor-changes.txt
new file mode 100644
index 0000000..e9df8eb
--- /dev/null
+++ b/doc/visitor-changes.txt
@@ -0,0 +1,7 @@
+Changes in visitor branch:
+o import_dxf(): layername="" imports all layers. Importing a single layer with a zero-length name is no longer supported. FIXME: The same prob. goes for dims
+o cylinder(): the r parameter will now always be used in place of a missing r1 or r2
+o for():
+  - with scalar argument now works, e.g.: for (i=23) echo(i)
+  - empty for loop is not evaluated, e.g. for () echo(i) 
+  - for loop with illegal value is not evaluated, e.g. for ([0:true:2])