Problem: From the point of view of the STEP format (and others), it is allowed
to describe a face on a surface with natural boundaries (torus, sphere) without
specifying these boundaries. Thus, a face on a closed surface and containing
an inner wire (or several) is correctly defined and describes a face with a
cutout defined by this wire. At the same time, there is a function
(ShapeFix_Face::FixOrientation) in the ShapeHealing procedure that corrects the
orientation of the wires, and it starts before the function of adding natural
boundaries (ShapeFix_Face::FixAddNaturalBound). There are many shapes that have
incorrectly oriented wires and this procedure successfully heals them, but on a
correctly specified face with single inner wire on closed surface, we do not get
the entire surface with a cutout, but a part of the surface defined by the wire.
This fix is intended to resolve this ambiguity.
Change:
1. Added function isNeedAddNaturalBound that returns TRUE if face needs to add
natural bounds.
2. Corrected condition in FixOrientation to ignoring faces that needs to add
natural bounds.
3. For tests in which one wire was incorrectly oriented on a closed surface,
flag AddNaturalBound was disabled.
5. Test with cutout from torus was created: bugs step bug28414.
Result: By default, it is correct to add natural boundaries, because this case
is correct from the point of view of the STEP format and others.
Possibility for adding LengthUnit info to XCAF document using special class XCAFDoc_LenghtUnit and XCAFDoc_LenghtUnitTool is implemented.
Package UnitsMethods is split: geom methods were placed to new file GeomConvert_Units which is in the toolkit TKXSBase, internal step scale factors was placed to StepData.
Updated UnitMethods to convert scale factor to different unit types.
Now, XSAlgo::XSAlgo_AlgoContainer is used to update unit info from static interface values.
New Draw command "XSetLengthUnit" and "XGetLengthUnit" for set or get XDE attribute.
Upgraded tests for STEP, IGES, OBJ, glTF, VRML formats to check area regressing with used unit.
Upgraded tests\de test cases to use any units in the "loop back" algorithms.
The "end" file from folder "de" has been corrected to exclude comparing the number of edges in the shapes from current result with test data.
"TODO messages generated for differences were removed.
Data exchange tests were regenerated.
Test parse rules are improved to recognize messages with word 'fail' or 'failure' as errors.
Test cases are corrected accordingly (TODO or REQUIRED statements added where fail messages were not noticed previously).
OSD::SetSignal() called with Standard_False argument to have FPEs disabled.
Test cased modification:
de step_4 G7 - fixed reference data.
de iges_2 H9 - Fixed reference data. According to input file, there are big tolerance edge and surface, which leads to big tolerance in ref data, result and image has changed due to usage of these shapes instead of omitting them.
de end - changed behavior in case of big isolines, now isolines not break fit command in certain places.
Modeling algorithms:
Fixed 0.0 / 0.0 uncertainty in circle to quasi-angular bspline conversion.
Correction of test cases for issue CR24589
Loop back tests for iges and step files with possibility of generation of reference data.
Clean up of test scripts (removal of obsolete code, adding comments, corrected formatting)
two fixes were done (same like bug 23722 and bug 23736)
Cancelling checking for number of warnings
rewrite some test scripts in cause of this and fix 23794
Tolerance checking was weakened.
Fix to check instability on Linux (bug 23822) was added.
TODO messages was added to tests which have different results on Windows and Mandriva2010
Ability to update this TODO in regeneration mode was added.
Add fix to bug 23852
Test groups were recomposed.
'dataexch' was renamed to 'de'.
CPU limit was decreased from 2000 sec. to 1500 sec.
