ReWork UpdateLoad functionality:
UpdateLoad calls only if provider is found.
UpdateLoad calls only for chosen type of translation (read/write)
Add new method to UpdateLoad all registered nodes.
Exception no longer occurs in current version. However, fuzzyvalue should be set to 1 for the result of the operation to make sense (result of common is 1 wire). Test case added
test bugs modalg_7 bug22821 failed on fillet shape with 5-th edge.
Experimentally has been found that reducing of parameter
ChFi3d_Builder::tolesp for this task solves the issue.
So, as soluton it is proposed to link parameter toleps with parameter range of
spine curve. So, production coefficient has been set to pass all tests and
2 teset was extended: tests/blend/complex/A6, tests/bugs/modalg_7/bug22821
first has been extended to test different scaling factors,
second has been extended to make fillet on all edges from 12.
Additionally:
- fixed misusage of tolesp in contexts where tolerance of point in 3d is excepted;
In some context usage of tol_esp is irrelevant, because its essentiality - tolerance of the parameter on the 3d curve.
So, in such context it has been replaced with new parameter tol3d (with fix value 1.0e-4).
Get rid of tolapp3d duplication constant - tol_3d
- tolesp = 5.0e-5 * (umax - umin)
- tolesp replaced by tolpoint2d/tolpoint3d in several classes.
Blend_Walking
BRepBlend_SurfRstLineBuilder
BRepBlend_RstRstLineBuilder
Blend_CSWalking
Instead `tolesp` - `tolgui` is employed in contexts where tolerance of guide curve parameter is excepted.
Instead `tolesp` - `tolpoint2d` or `tolpoint3d` is employed in contexts where tolerance of point in 2d or 3d space is excepted.
- Replace tolesp with tolpoint2d/tolpoint3d in BBPP function argument.
- Use tolapp3d instead tolesp in BonVoisin function,
distmini of two edges returns two solution points instead one.
Second unneeded solution point is the same as first.
The problem was in fact that second edge has continuity C0.
In this case additional extremas analysis performed in special procedure
PERFORM_C0
And second point found in this procedure.
Folowing code of
BRepExtrema_DistanceSS::Perform (variant for Edge/Edge)
we should this additional solution extremas set
to be object of test TRI_SOLUTION before pushing
into main list of solution.
This solves the problem.
Corresponding test and compound with edges has been created.
Additionally, in the function
PERFORM_C0
an obvious error has fixed.
Problem: The shape has two faces that dimensions are less than tolerance in the specified STEP file. This situation produces the invalid shape.
Change: Shape process has operator FixFaceSize, that is exactly suitable for this shape. The only option of FixFaceSize operator was added to InterfaceStatic, to manage the operator from DRAW command.
Result: The shape without artifacts can be reproduced from STEP file with an extra healing procedure.
BSplCLib_2.cxx - method MergeBSplineKnots is modified in order to have always not empty result.
Geom2dConvert.cxx - in static function MultNumandDenom tolerance for comparing knots is decreased.
tests/bugs/modalg_8/bug33179 test case added
BRepOffset_MakeOffset.cxx - tolerance control for building planar faces is implemented,
updating tolerance for initial entities is avoided
BRepAlgo_Loop.cxx - "total" setting tolerance 0.001 is removing
BRepLib.cxx - checking of "locked" for vertex is removing in static function UpdShTol
QABugs_20.cxx - add new command OCC26441 for checking tolerance differenses between two "identical" shapes
tests/bugs/modalg_7/bug30054 - case now is "BAD", because really result shape is invalid: many faces has not closed wires with huge gaps between ends of edges. Result was "OK" only because tolerances of vertices were increased by algorithm to cover all gaps.
tests/bugs/modalg_8/bug26441 - new test case added
Other test: B3, C8, A7, C8: they were "BAD" and now are "BAD", only some problems are changed.
Problem: Desynchronization of behaviors of GeomAdaptor_Curve::NbIntervals and
GeomAdaptor_Curve::Intervals functions. First calculates number of intervals, then
array is created and second fills the array. In some cases the size of array
is less than need for filling.
Change:
1. Added function BSplCLib::Intervals that calculates number of interval and fills
the array with its (if needed).
2. Simplified the algorithm of intervals calculation.
3. GeomAdaptor_Curve::NbIntervals/Intervals and Geom2dAdaptor_Curve::NbIntervals/Intervals
use BSplCLib::Intervals.
4. When creating an adapter for the base curve, the boundaries of the adapter for the offset curve are applied.
5. Test for problem shape was created: bugs modalg_8 bug33187.
Result: The new approach eliminates the problem of writing outside the array bounds.
Added step of refinement the coarser of the two shapes meshes to produce two meshes with approximately the same density.
Added tests lowalgos/proximity.
Fixed accounting of parameters to adjust number of initial sample points
GeomLib_IsPlanarSurface.cxx - using poles for checking BSpline, Bezier curves and surface changed
on checking by curve, surface points.
BRepOffset_MakeOffset.cxx - set normal of plane surface according to normal of initial face surface
tests/cr/bugs/bug33170 - new test case added
Added new CMake variables to build using DRACO
Windows OS works with Debug and Release configuration DRACO
Linux works with Release configuration DRACO
Fix warnings:
* catching polymorphic type 'class Standard_Failure' by value [-Wcatch-value=]
* this 'if' clause does not guard... [-Wmisleading-indentation]
ShapeAnalysis::OuterWire(): fixed missed logic when TopoDS_Iterator notifies about more objects to iterate, but there are only vertices and no additional wires at all.
Add parameter EnableControlSurfaceDeflectionAllSurfaces to IMeshTools_Parameters enabling possibility to optimize mesh even on analytical surfaces;
Add corresponding parameter -surf_def_all to incmesh Draw command.
Method IntWalk_PWalking::ExtendLineInCommonZone(...) now is not called if the already found intersection point is on surface boundary.
As result, the intersection line going along any boundary will never be extended. It is appropriate for high-level OCCT-algorithm because they will take the boundary (not computed line) as intersection result.
Increase minimum number of discretization points by one explicitly on each iteration of model healer to cover cases degenerated to line (for cases when face consists of 2 edges only).
Fixed the Toroidal surface recognition;
The test case bugs/moddata_1/bug22296: the result is a torus with parameters: u [0, 2*PI], v [PI, 3*PI];
The test case bugs/modalg_5/bug23954: BAD -> OK.
With enabled flag BUILD_USE_PCH we get an error of compilation of TKService:
1>Image_VideoRecorder.obj :
error LNK2019:
unresolved external symbol
"int __cdecl av_strerror(int,char *,unsigned __int64)"
(?av_strerror@@YAHHPEAD_K@Z) referenced in function
"protected: class TCollection_AsciiString __cdecl Image_VideoRecorder::formatAvError(int)const "
(?formatAvError@Image_VideoRecorder@@IEBA?AVTCollection_AsciiString@@H@Z)
1>Media_FormatContext.obj :
error LNK2001:
unresolved external symbol
"int __cdecl av_strerror(int,char *,unsigned __int64)" (?av_strerror@@YAHHPEAD_K@Z)
And many other similar errors. Similar errors occures in some other projects too:
TKService
TKOpenGl
TKOpenGles
TKXCAF
TKXDEDRAW
TKDFBrowser
TKMessageModel
TKMessageView
TKShapeView
TKTInspector
TKTreeModel
TKVInspector
TKView
Proposed solution: turn off cotire from targets, whose compilation cause error while applying cotire tool.
In this ticket migration to cotire 1.8.1 (from 1.7.9) is done.
COTIRE_PREFIX_HEADER_IGNORE_PATH does not apply here, because its employing
causes some errors in 3rdparty libraries (for example, in TKService project).
Projects (TKDFBrowser TKMessageModel TKMessageView TKShapeView TKTInspector TKTreeModel TKVInspector TKView)
which use Qt may be proceeded by cotire tool, but after fixing a bug of cotire:
https://gitlab.kitware.com/cmake/cmake/-/issues/18353
0033153: Fix cotire bug, causing impossibility of compilation targets depending from Qt.
Solution is proposed in:
https://gitlab.kitware.com/cmake/cmake/-/issues/18353
After emploing intel oneTBB of version 2021.5 on NEXT platforms arises error on OCC Pr.:
`
*/ld: */libfreetype.a(truetype.o): relocation R_X86_64_PC32 against symbol `TT_RunIns' can not be used when making a shared object; recompile with -fPIC
*/ld: final link failed: bad value
`
Proposed solution: linking with shared object of libfreetype, rather than with archive library.
Despite that error occurred only in OCC Pr., it is worth to accept changes for OCCT too.
Previous approach do not taking into account, that oneTBB library may be installed in the system folder
on Linux. So, solve the problem it is proposed to employ find_package utility, that works great
as on Windows, as on Linux.
