LICENSE:
     All patches are licensed under original Open CASCADE license



Installation instructions:

- 0. You are expected to apply some efforts to install these patches. They are source code patches only but to apply them you may
     need to generate some additonal files (e.g. *.hxx) and modify project files. Volunteer help to complete these patches to
     make them easily install-able over Open CASCADE installation would be much appreciated.
- 1. Fixes must be installed in ascending order (subversion xxx.y must be installed over xxx) into $CASROOT/src directory.
     _baseline subdirectory is only for reference and should be ignored.
- 2. Contents of one fix can overwrite previous fix(es).
- 3. Modifications of *.cdl, *.gxx require using WOK (as this may generate files in inc and drv directories).
- 4. Addition or removal of files require modification of project files (e.g. *.vcproj) to take them into account.
- 5. Source code was modified on Windows. There is a high chance that it contains Windows or mixed line endings (0x0D 0x0A).
- 6. Though certain amount of testing has been performed to verify correctness of the fixes and eliminate probability of
     regressions, such risk may exist.


NOTE: As of fix380 details are given in the reversed chronological order

fixes till fix405 (included) released as 6.3.1 Update 6
=======================================================

fix405
- Summary: Corrected processing of parameter read.iges.bspline.continuity set to 0 
- Detailed description: When read.iges.bspline.continuity set to 0 (with Interface_Static::SetCVal()), the BSplineSurface
	is not enforced to become periodic if it is closed what is done when it's set to 1. In some cases (e.g.
        www.firstcadlibrary.com/seatmotor2.igs, entity 397) this leads to wrong resulting shape after Shape Healing. Instead
	of added missing seam the face is split into a compound of 2 invalid faces.

	NOTE: See comments in the code, for consistency between TransferSplineSurface() and TransferBSplineSurface() the
	former must have a code to convert to periodic surfaces, and the latter - to IGESConvGeom for enforcing continuity.
	It appeared that OCC IGES converter never enforced surface continuity though it was expected to. It is up to the OCC
	team to verify if addition of these calls will not break regressions.

- Dependency: None

fix400
- Summary: Fix uninitialized memory access issues 
- Detailed description: The code compiled on Linux with gcc4.4.1 in release mode produced hang on GeomPlate algo while
                        worked in debug mode. Analysis with Intel Parallel Inspector revealed several uninitialized memory
                        access problems. This fix addresses them
	- IntRes2d_Transition, _Domain; Intf_SectionPoint; Adapto3d_CurveOnSurface; Bnd_Box2d - initialization of all fields
- Dependency: None

fix395
- Summary: A set of fixes to enable thread-safety
- Detailed description:
	- Plate_Plate now uses field members instead of static variables
		- U2, R and L should be local variables by their meaning. However, there are some corner cases in 6.3.1
		 implementation when they should inherit their values from Uold, Vold, not U,V. I preserved current behavior by
		 using member fields, but with deep code review this can be changed by OCC team. 
	- ProjLib_CompProjectedCurve, Approx and BRepAlgo_NormalProjection now use chronometers in a special compile mode (not in plain debug)
- Dependency: fix265

fix390
- Summary: Support of user-defined memory manager
- Detailed description: Using MMGT_LIBRARY environment variable, a user-defined memory manager can be dynamically loaded
	during load time
- Dependency: None


fixes till fix385 (included) released as 6.3.1 Update 5
=======================================================


fix385
- Summary: Port on Linux and MacOS
- Detailed description:
	- Makefile.* in ./make/TKernel/ and TKDraw
	- AdvApp2Var_ApproxF2var.cxx uses Max() and Min() instead of max(), min().
        - NCollection_BaseAllocator::CommonAllocator() avoids using Standard_Mutex::Sentry as it uses
          static Standard_Mutex (see Standard_ErrorHandler.cxx) which may be uninitialized yet
	- Standard_Atomic.hxx : atomic operations for MacOS are the same as for Linux
	- AdvApp2Var: renamed macros abs, min, max to advapp_* to avoid name conflicts on Linux/Mac which forces
          gcc to use standard C functions and therefore run-time fails (e.g. integer abs(0.92) on Mac will return
          integer 0, not double 0.92).
- Dependency: All previous fixes, as this one is just a port of update 4 to other platforms


(DEFERRED) fix380
- Summary: Replace FILE* interface of LDOM_XmlWriter with ostream&
- Detailed description: The API that accepts FILE* causes crash when mixing VC run-time libraries (i.e. OCC was compiled
	with VS2003 and the application that creates FILE object with VS2005. There are different sets of file
	handles in each run-time and therefore VS2003 incorrectly processes objects created with VS2005.
	New API accepts file name so that the file object is managed by single run-time.
	See my blog post in March 2010 on opencascade.blogspot.com
- Dependency: None


=======================================================





List of fixes

Open CASCADE 6.3.0
==================


fix101
- Summary: Thread-safe type instantiation
- Detailed description: *.edl and Standard_DefineHandle.hxx are modified to use Standard_Mutex to protect
      STANDARD_TYPE() object instantiation. 
      *.edl were modified using OCC 6.3.0 version not in wok/lib. Some files were different and perhaps 
      requires further revision by OCC team.

      WOKTclTools modified to use auto-generated WOKTclTools_Intepretor file instead of hard-coded. OCC team should
      physically remove WOKTclTools_Interpretor_proto.hxx and WOKTclTools_Replace.tcl files from the directory.
   
      Additional improvements in CDL extractor and Standard_DefineHandle.hxx:
      - removed redundant Standard_EXPORT in _Type() function;
      - _Type_() made to return const&, not &
      - DownCast() made to return Handle(), not const Handle() which is redundant
      - removed "// Methods PUBLIC" and alike as they had no value add
      - 0 parents in _Type() (1 used before was incorrect but never used in OCC);
      - using Handle() and STANDARD_TYPE() macros everywhere (instead of Handle_ and _Type) function for better consistency;
      - minimized used of #ifndef _*_HeaderFile in callers for better readability;
      - add virtual to DynamicType()
      - removed code duplication in Standard_DefineHandle - use of common macros
      - removed redundant second argument in IMPLEMENT_STANDARD_RTTI macro - this potentially may break source compatibility!
      - CDL extraction uses DefineHandle macros instead of own generation
      - (NOT DONE) DEFINE_STANDARD_RTTI: to remove redundant argument - will break source compatibility!
      - (NOT DONE) make Handle_*.hxx empty (or even remove at all) puting everything ito *.hxx with help of macros. This will
        require C++ code change of CDL extractor as it creates #include <Handle_*.hxx> directives
      - (NOT DONE) cleanup .jxx if possible

  - todo
    - make Standard_Transient (Persistent)::Delete non-virtual to ensure that object is really destroyed in
      Handle::EndScope(). This requires review of its ancestors.
    - (NOT POSSIBLE) put pragma warning into IMPLEMENT TYPE
    - remove Destroy() and destructors in descendants of Standard_Transient (e.g. ~Standard_Failure(), WOKTclTools_Interpretor, etc)
    - Update Handle_*_proto.hxx to use macros
    - Downcast to macro and into Handle_ and IsKind() inline (ABV's suggestion)
    - modify CPPClient_Template.edl, CPPExt_TemplateOBJS.edl, _OBJY.edl, CPPIntExt/Engine_Template.edl, 
      CPPIntExt/Interface_Template.edl to match templates generated by CPPExt and CSFDSchema

- Dependency: None.
- Files:


fix105
- Summary: Streamlining root Handle class files (Standard_Transient)
- Detailed description:
	- (NOT POSSIBLE) removed Standard_Transient.hxx in src directory, Standard_Transient_proto.hxx is used
       	  to generate it (similar to _Persistent_proto.hxx). OCC team should physically remove
          Standard_Transient.hxx in src.
	- removed redundant macro NullObject_Raise_if in Standard_Transient_proto.hxx
	- implemented _Type() via macros IMPLEMENT_STANDARD_TYPE
	- more Standard_Transient methods made inline

  - todo
    - do similar modifications in Standard_Persistent. Note that Persistent is not required to be thread-safe, as
      data storage/retrieveing is anyway sequential
    - //remove Delete() and destructors in descendants of Standard_Transient (e.g. ~Standard_Failure(), WOKTclTools_Interpretor, etc)
    - Downcast to macro and into Handle_ and IsKind() inline (ABV's suggestion)
    - Handle Make Handle_Standard_Persistent similar to Handle (do not use IsReentrant() - must be documented)
    - check Destroy() to be virtual 

- Dependency: fix101
- Files:

fix106
- Summary: Miscellaneous OCC improvements, including improved thread-safety
- Detailed description:
	- removed OSD_Host from IGESData.cxx and IFSelect_Functions.cxx. OSD_Host calls ::gethostbyname() deprecated
	  function that downgrades performance (also reported on OCC forum), creates an orphan thread that skews
   	  concurrency analysis. Anyway its results are not used (see commented code in IGESData.cxx)
	- BSplCLib and BSplCLib - removed static variables thereby making files thread-safe. defined M_SQRT2 as
	  constant (see math.h), not as expression sqrt (2.). This will help performance.
	- Standard_CString.hxx ::IsEqual(strings) corrected.
	- Transfer_Process:
		- removed self-made "optimization" (storing last used key and binder) which is
	          non-applicable in multi-threaded environment
		- FindAndMask() removed
		- Some methods made returning const& (in fact, CDL contained errors in methods declarations)
		- miscellaneous code optimizations
fix106.1 (see also fix204)
	- removed all new/delete (as these are expensive) and replaced with auto-allocation (on stack) per ABV
	  suggestion on my blog
	


- Dependency: None. Can be safely merged with fix105 and/or fix110.
- Files:

fix110
- Summary: Modified files to be more reliable regarding including windows.h in other header files
- Detailed description: windows.h can be included via some Open CASCADE *.hxx files (e.g. in Standard) and this may
	affect function and variable names in .cxx files (e.g. var name matching macro defined in windows.h).
	Code modified to use unique identifiers or to include windows.h in the appropriate order
- Dependency: None.
- Files:

fix112:
- Summary: Improvement of Standard_Mutex 
- Detailed description: 
	- Standard_Mutex:
		- Standard_Mutex.hxx and .cxx modified to remove windows.h from *.hxx (to avoid spoiling files
		  including it where order of including windows.h is important). windows.h breaks compilability
		  of some .cxx files (including auto-generated) being inserted in the beginning before the file
		  gets a chance to define its own windows.h-related macros.
		- implementation is hidden into .cxx (due to windows.h independence)
		- Lock() is implemented via standard EnterCriticalSection() instead of TryEnter...() + Sleep(). This
		  is required for effective multi-threaded applications (proven via analysis with Intel Thread 
		  Profiler and Parallel Amplifier). 
		- Further possible improvements - add spincount as constructor argument and call 
		  InitCriticalSectionAndSpinCount.

- Dependency:
- Files:

fix115: (experimental!!! do not include into production code)
- Summary: Providing framework for executing OCC algorithms in parallel
- Detailed description: 
	- OSD classes that enable execution via pipelines running in parallel.
	- todo 
		- OSD_Pipeline should create threads with OSD::SetSignal(val) - with val equal to master
		  thread. Is there are a way to get it ?
		- OSD_Scheduler::PerformParallelLoop() can be tried with TBB parallel_for.
                - add Long to Standard and replace in OSD_WaitCondition::wait. This is important !!!
                - OSD_WaitCondition and _InverseSemaphore must be ported to Unix

- Dependency: fix101, fix112.
- Files:

fix116
- Summary: Improving thread-safety
- Detailed description: 
	- Management of static objects in gp and Message to avoid data race (i.e. access to unprotected data
	  from different threads)
- Dependency: None.
- Files: 

fix117
- Summary: Corrected implementation of OSD_InverseSemaphore - as object manipulated by value, not by handle
- Detailed description: 
	- InverseSemaphore must not be handle because Handle_Standard_Transient is not thread-safe and
		 ISemaphore is used across threads. Therefore used as a pointer.
     OSD_Pipeline:
	- OSD::SetSignal() is now added to every thread that is created.
	
- Dependency: fix115.

fix118
- Summary: Improved exception handling
- Detailed description: 
	- Exception objects (Standard_Failure ancestors) are now defined/implemented using macros (see Standard_DefineException.hxx)
	  in a way similar to defining handles. This enables creating user-defined exceptions without CDL.
	  Note that as preprocessor does not allow to insert preprocessor instructions (starting with #) into macro
	  definitions, it is impossible to define macros #<object_type>_Raise_if(). Therefore their generation is
	  kept within CDL extractor as previously. The user does not really need to do this - he may rely on No_Exceptions
	  macro. CPPExt_Template.edl must be copied to $WOKHOME/lib for CDL extractor to take effect.
	- Standard_Failure has Throw() non-virtual and ancestros does not have it anymore (implementation was the same).
    
     OSD_signal_WNT.cxx
	- OSD::SetSignal() now always sets signal handlers, exception handlers and structured exception (SE) translators. This is important
	  to deal correctly in case if user's code is compiled with /EHs or /EHa. UseSETranslation() are removed,
	  as redundant. See detailed documentation in OSD_signal_WNT.cxx on OSD::SetSignal().
	- very detailed documentation for OSD::Setsignal() in Doygen format is now in this file.
        - removed static variables that set signals only once to enable calling ::SetSignal() for other threads or even
	  within the same thread if the user switched handlers before
	- mutexes are used to control multi-threading.


	- TODO:
		- Minimize signal handling by OSD_Signal.cxx to SIGSEGV and SIGFPE. On Unix it currently processes
		  all (or almost all) signals (e.g. SIGHUP) by exiting. There is no value add to that. It's better to be
 		  done by user apps.
		  OCC should only handle those which it can survive after, if at all it must do anything. OCC classes may
		  or maynot be provided for convenience but should not be set in the OCC itself.
		- OSD_signal.cxx - if possible, get rid of sigaction() and replace with signal() and and other more
		  advanced staff. We just need to map signals to C++ exceptions, so let's not make code more complex than needed. This
		  can simplify the OSD API and remove. If possible, minimize implementation discrepancies between OSes.
		- On SGI - remove use of env variable TRAP_FPE and replace with function handle_sigfpes() (as this is more
		  reliable) - see SGI doc (http://techpubs.sgi.com/library/tpl/cgi-bin/getdoc.cgi?coll=0530&db=man&fname=/usr/share/catman/p_man/cat3/ftn/fsigfpe.z)
		- minimize exposing OSD API by removing SegvHandler(), WntHandler() and others and keep SetSignal() only.
		  Alternative approach is instead expose functions which map to OCC exceptions and document them, so that
		  the user can reuse them in his own handlers.
		- rename OSD::SetSignal() with SetHandlers() as it not only sets signal handlers but exceptions also.
		- removed message box and call to a debugger. This is an ugly way and only pollutes the code. Moreover call
		  to Debugger does really work on modern Windows (debuger is still ironically called due some failure within
		  OSD_signal_WNT.cxx). In reality debugger can be attached to the process. Moreover, CSF_DEBUG_MODE was
		  seemingly never tested - the message box could not display for FPE signals (because _fpreset() and
		  _clearfp() were not called before it). In the ultimate case this code can be moved to some DRAW function
                  as an example of how to implement own signal handling.
		- add additional boolean argument Reenforce into OSD::SetSignal() that would not redefine already set user's
		  defined handlers. In this case, OSD_Pipeline should call OSD::SetSignal(aFloatingSignal=true,
		  aReenforce=false) and would not reset user's handlers if they were redefined by user in the main thread

	
- Dependency: fix106, fix117.

fix118.1
- Summary: Corrected fix118 by adding modified OSD.cdl
- Detailed description: Removed UseSETranslator() methods.
- Dependency: fix106, fix117, fix118.

fix119
- Summary: Part of fix120 (done after it) - simplification of IGESToBRep_CurveAndSurface
- Detailed description: TransferGeometry() streamlined.
	- No GroupWithoutBackPointers() - never called due to IsKind()
	- transformation always attached
	- Regressions:
		- BUC60743 - wrong positions of free curves (due to double application of transformations -
		  in TransferBasicCurve() and in TransferGeometry()). 
		- Warnings for subfigure and group entities failed are no longer issued (can be fixed) - BUC60743
		- double registration on the map for TrimmedSurfaces (in TransferGeom() and TransferBasicSurface())
- Note: Will roll back and address later. Note that OCC6.3.0 has potential bug - not applying locations for
		  trimmed surfaces (see for_bugs/seatmotor2-modified.igs and its boundary surface DE3).
                  In 6.3.0 locations of trimmed surfaces (and perhaps some other faces) are not taken into account.
                  See IGESToBRep_CurveAndSurface::TransferGeometry() - it just calls TransferTopoSurface() which 
                  does not apply transformation to TrimmedSurfaces. If to add transformations into TransferGeometry()
                  then a bug with BUC60743.igs appears - because free curves are moved already and there becomes
                  double move. The wayout is to significantly rewrite IGESToBRep - either applying transformations
                  individually (e.g. creating a protected method in _CurveAndSurface and calling from everywhere 
                  where TransformedXXX() methods are not called) or extend fix119 and always call TransferGeometry()
                  but this may impact performance due to if's(). Another way out is to create drivers for each type 
                  of entities (what I did in X3D) and call TransferGeometry(). The downside of solution #2 and #3 
                  - is that size of the shape can be bigger due to attached locations instead of directly modified
                  definitions.
- Dependency: None

fix120 (experimental!!! do not include into production code)
- Summary: Making translation of IGES groups (type 402) parallel
- Detailed description:
	- IGESToBRep_CurveAndSurface:
		- access to Transfer_Process protected with Mutex.
		- note that IGESToBRep_CAS made not trully effective with static mutex. In a better case it should 
		  be a field, but the issue is that mutex should not be allowed for copying (as CriticalSection 
		  forbids that).
	- todo 
		- MoniMessage_ProgressIndicator must be rewritten for concurrent applications and added to
		  IGESToBRep_CAS

	
- Dependency: fix106, fix101, fix115, fix116.
- Files: 


fix202:
- Summary: Fix regression to enable display of OCAF-stored shapes in 3D view in DRAW + commands for setting background
- Detailed description: OCC 6.3.0 comes with ViewerTest tailored to use AIS_InteractiveContext with NIS_View.
	Trying to display OCAF-stored shapes in new 3D views (e.g. with XShow command) lead to crash due to
	DownCast() to NIS_View which resulted in null handle.
	ViewerTest_ViewerCommands.cxx extended to set background (color and gradient color)
- Dependency: None


fix203
- Summary: Improving BOP performance
- Detailed description: Redesign of IntPolyh 
	- one generic Array and instance of all others
	- Init with minimum number of elements and resize in the future if required
	- Standard::Allocate() instead of new, as default values are not used
	- IntPolyh_ArrayOfStartPoints removed as it was out of use (MaillageAffinage did not implement GetSP() method). 
		Use IntPolyh_ArrayOfTangentZones instead.
- Dependency: None
- See also:	fix375


fix204
- Summary: Creation of OSD_ReadWriteLock and improvement of BSplSLib, BSplCLib and PLib to be usable for multi-threading with better performance
- Detailed description: OSD_ReadWriteLock is implementation of a patter readwritelock (see wikipedia for example) that
		allows multiple concurrent reading but once there is a writer it blocks other writers and readers
		*Lib are now use TLS (thread-local storage) instead of constant allocation/deallocation which is time-consuming
		With this implementation *Lib are slightly less fast than in 6.3.0 but much faster than in initial fix106 (where
		allocation/deallocation was used)
	- REMAINS TO BE COMPLETED OR 
		- OSD_RWL can be used in IGESToBRep_CurveAndSurface when accessing its map, to increase concurrency
		- ABV's idea to use auto variables in *PLib or avoid using RWL creating buffers only once
- Dependency: fix115.1, fix115

fix210:
- Summary: Extrema redesign
- Detailed description:
	- Temporary modification: GeometryTest
fix210.0:
	- IntTools_BeanBeanIntersector: 
		ComputeUsingExtrema() - take one of two GeomAdaptors out of cycle as it does not change
	- Extrema
		- SquareDistance. Check TrimmedSqDistances !! Search Distance(), Magnitude() inside Extrema
		- Point return const& and update callers (may add private copy constructor and operator= to ensure update of all callers)
		- Loading and deferred perform
		- caching point calculation
		- Raise_if instead of ::Raise()
	- Updated callers: gce_MakeCirc, GC_MakeArcOfCircle, etc
fix210.1:
	- Extrema:
		- Replacement of all internal calculations with SquareDistance
fix210.2:
	- IntTools:
		- create adaptor only when needed but still once per UsingExtrema() call
fix210.3:
	- Extrema:
		- ExtCC, _GenExtCC, _GExtCC:
			- deferred perform, independent loading of curve, caching of points, etc
			- simplified code of GExtCC::Perform()
fix210.4:
	- Extrema:
		- _CurveCache and switching _GExtCC and _GenExtCC to its using for maximum reuse of calculations
fix210.5:
	- Extrema:
		- cache of myDu and myDv in FuncExtCC

	- TODO (if needed):
		rename into SquareDistance() and LowerSquareDistance(), this may require further updates downstream and update of documentation
		Geom2dAPI_*,
		GeomAPI_*,
		GeomFill_SectionPlacement
		BRepExtrema_DistShapeShape

- Dependency: None



fix211
- Summary: Speed up of Geom_BSplineCurve, _BSplineSurface and Geom2d_BSplineCurve ::IsCacheValid()
- Detailed description: Division NewParameter = (NewParameter - parametercache) / spanlenghtcache  takes much time, so normalization is avoided
- Dependency: None

fix212
- Summary: Speed up of math_FunctionSetRoot (used in Extrema)
- Detailed description: 
	- Excluding recalculation in points which did not progress during iteration 
	- Sqrt replaced with sqrt
	- minimized use of sqrt
	- static variable mydebug use surrounded with #ifdef DEB
	- minimized use of Abs in Bounds()
- Dependency: None

fix215
- Summary: Declaring methods as protected in XCAFApp_Application and XCAFPrs_AISObject
- Detailed description: Declaring protected instead of private enables subclassing these classes
- Dependency: None

fix220
- Summary: Removal of extra OpenGL buffer clearing to minimize flicker
- Detailed description: Inspired by fix on http://www.opencascade.org/org/forum/thread_8291/. Extended to cover
	 non-Windows platform (not tested !)
- Dependency: None

fix225
- Summary: Support of progress indicator in RWStl and optimization of reading Ascii.
- Detailed description: 
	- Progress indicator added to Write...(), Read...(). 
	- reduction of string buffers reallocation when writing STL files in ascii mode
- Dependency: None

fix230
- Summary: BRepTools_ReShape::Status() incorrectly returns new result for shapes with location
- Detailed description: With MyConsiderLocation mode = True, Status() returns a shape with null location
	Fixed version now uses Value() to check the result. Additional fix in Value() - to multiply result
	location by shape location, not vice versa.
	Test case - import shaver.igs into XDE, and check that subshapes are of full yellow color. Quick
	check - entity 10921 and its 1st face (10922). Prior to the fix, the TransferProcess incorrectly
	stored a null location result and therefore colors were not assigned to the resulting subshape.
- Dependency: None


fix235
- Summary: Incorrect placement of ZBuffer trihedron
- Detailed description: OpenGl_triedron.c used weird check isWithinView and correctly placed ZBuffer trihedron
	only when the model went out of the view. Fix ignores this check. Sergey Anikin claimed this must have
	been fixed in 6.3.1.
- Dependency: None

fix240
- Summary: Minor performance improvements
- Detailed description: Variables out of cycle, use of const& instead of =.
- Dependency: None

fix250
- Summary: Caching of B-Splines
- Detailed description:
	- Caching at the last span never worked - it stored incorrect span length (negative) and made wrong checks 
	at IsCacheValid()
	- (Not fixed yet) BSPlineSurface called BuildCache() at middle of a span being cached, while curve -
	  at start parameter of a span. why ?
	- Curve uses cache up to D3(), Surface - D2 
- Dependency: None

fix255 (Deferred) 
- Summary: OSD_Chronometer (and _Timer) to be based on GetProcessTimes() to correctly measure time in multi-threaded apps
- Detailed description:
- Dependency: None

fix260
- Summary: BSplCLib, BSplSLib, PLib - made reentrant 
- Detailed description: Use of automemory (on stack) and heap only if default size (for 4096 elements, i.e. 32K) is not enough.
	fix106.x, fix204 must disregarded. 
- Dependency: None


fix265
- Summary: Thread-safety/re-entrability improvements
- Detailed description: 
	GeomConvert_ApproxCurve/_Surface - protect shared resource with mutex (cannot make shared resource local)
	IntCurve_IntPolyPolygen.gxx - make variable local (though it's out of use effectively)
	Intf_* - make static vars either const for explicity, or class members, or local variables; for some class this required to move static function as private method
	IntcurvesFace - make static variables local
	IntCurveSurface - make static variables local and const where possible
	BRep_Tool - local instead of static
 (+fixes for UVBox)
	IntImp - use of local buffers and wrapping with arrays (to avoid heap allocation and adverse impact on performance)
	BRepLib - local instead of static, using mutex
	ProjLib - made static var a class field
	
	Approx - made static var a class field, use of mutex (like in GeomConvert_Approx)
	BRepclass3d - local instead of static; _SolidExplorer.cxx - simplified code, made static var a class field, local instead of static
 Everywhere to use Sentry !!!
	
	
- Dependency: None

fix267
- Summary: ShapeExtend_WireData::Edge() to return const& 
- Detailed description: This helps performance due to reduced data copying and handle::BeginScope(),::EndScope()
- Dependency: None


fix270
- Summary: Typo in ShapeFix_Wire::FixEdgeCurves() for checking same range
- Detailed description: Note that ShapeAnalysis_Edge::PCurve() returns reversed range for reversed edge
	 while comparison is done regardless orientation
- Dependency: None


fix275
- Summary: Using higher precision (17) in BRepTools::Write()
- Detailed description:
- Dependency: None


fix280
- Summary: ShapeFix_Edge::SameParameter() now retains original 3D interval
- Detailed description: See comments ShapeFix_Edge::SameParameter(). In the case of periodic curve, 3D range lying
	outside of a period was previously adjusted to fit into it. Thus copyedge used different ranges than original
	edge and in the case its pcurves had to be copied back, ranges and pcurves became inconsistent.
	Test case - edge-fix280.brep. Use fixshape command in Draw and check 3rd edge. Withoutout the fix the pcurve
	became wrong and new redundant edges were inserted.
- Dependency: None

//fix285 (deferred)
- Summary: Enable subclassing Message_ProgressSentry making its fields protected and methods virtual
- Detailed description: Subclassing can be useful for example, for nested sentries for multi-threaded apps where methods will have to use
	 locks before calling Message
- Dependency: None


fix290
- Summary: Thread-safety/re-entrability improvements
- Detailed description:
	AIS_InteractiveContext.cxx, AIS_GlobalStatus.cxx - avoid using static ListIterator
- Dependency: None

fix295 (deferred)
- Summary: Make the code of the static function GetDeflection() in Prs3d_ShadedShape.gxx accessible as static method
- Detailed description: This is to avoid code duplication in user's code 
- Dependency: None

fix300
- Summary: Standard_MMgrRaw::ReAllocate() trying more if original realloc() fails
- Detailed description: If original realloc() fails ReAllocate() will now Free() and Allocate(). This can help in two cases:
	a. when free memory is low and freeing already allocated chunk can help get a larger chunk (e.g. if free chunk
  	   is adjacent to previously held
	b. when replacing OS memory allocator (e.g. with Intel TBB which initially discovered this issue) when
	   memory was allocated using OS malloc() and then substituted realloc() returning NULL causing Standard_OutOfMemory
	   exception.
        Note that in either case initially allocated memory will leak.

- Dependency: None


fix305
- Summary: Improvements in Vrml 2.0 exporter to make it 
- Detailed description:
	- correct processing of normals of reversed faces;
	- enabling customization of default materials;
	- making export of normals optional
	- possibility to customize WorldInfo
- Dependency: None

fix310
- Summary: Adding debug output when any Standard_Failure exception is thrown
- Detailed description: Standard_Failure::Throw() method has std::cerr output regardless if the exception will be caught
	 or not. This helps to notice exceptions that will eventually be caught.
- Dependency: None

fix315
- Summary: Corrections in BRepExtrema
- Detailed description: Added sqrt() to correct oversight in fix210 (noticed my Francois Louzon) 
- Dependency: None

fix320
- Summary: Corrections in VRML reader
- Detailed description: 
	- Parser buffer extended from 8K to 64K
	- Correct processing of string lists (MFString type)
- Dependency: None

fix325
- Summary: More corrections for thread safety
- Detailed description:
	- PLib 
- Dependency: None


Open CASCADE 6.3.1
==================

fix330
- Summary: StlMesh_Mesh now have protected (not private) data fields
- Detailed description:
	- StlMesh_Mesh assumes inheritance (many methods are virtual) but subclasses cannot access/change data fields
- Dependency: None

fix335
- Summary: Corrected modifications in fix265 in ProjLib to restore functions signature.
- Detailed description: Corrected modifications in fix265 in ProjLib where signature of methods was changed from const to non-const. This
	 hade the fact that virtual method were not overridden what caused exception when using ProjLib_CompProjectCurve
	 (in GeomPlate algorithm). Now the fix usis const_cast in run-time.
- Dependency: none (no direct dependence on fix265)

fix340
- Summary: Improving thread-safety
- Detailed description:
	- GeomFill.cxx - removed static variables
- Dependency: None

fix345
- Summary: BRepTools_ReShape now is able to accumulate shape mapping from another ShapeBuild_ReShape
- Detailed description: New method void BRepTools_ReShape::Join (const Handle(ShapeBuild_ReShape)&) added.
	The use case for it is concurrent processing of shapes and then reduction of results into a single ReShape.
- Dependency: None

fix350
- Summary: Improvement of thread-safety by adopting class _EvaluationFunction instead of a function.
- Detailed description:
	- Modifications analogous to ones introduced in 6.3.1 for AdvApprox et all.
	- BSplSLib, BSplCLib, AdvApp2Var_EvaluatorFunc2Var
	- GeomConvert (removes fix265 modifs), Geom2dConvert, GeomLib, GeomPlate

	AdvApp2Var contains a lot of Fortran-derived code which uses static variables. Conversion into class
	members will be a heavy and attention-required task. Until that done, any call to AdvApp2Var (or
	algorithms using it - such GeomApprox_Surface called from ShapeHealing BSplineRestriction or GeomPlate)
	require use of global lock.
- Dependency: None


fix355
- Summary: Improvement of BRepTools_Modifer
- Detailed description:
	 Removal of static in BRepTools_Modifier
	 Modifier lost pcurve ranges on non-SameRange edges
- Dependency: None

fix360
- Summary: (POSTPONED) Corrected processing of pcurves in BRepTools_TrsfModification::NewCurve2d()
- Detailed description:
	Redundant TrimmedCurve
	Trim for periodic curves
	PCurves on planes
- Dependency: None

fix365
- Summary: Extending XCAFPrs_AISObject to enable customization of default color
- Detailed description: virtual DefaultStyle() method added.
	Note that default style object has not been made a data member to avoid memory consumption
- Dependency: None


fixes till 365 released as 6.3.1 Update 2
=====================================================

fix370
- Summary: Optimization of BRepMesh_DiscretFactory to cache registered factory methods
- Detailed description: In 6.3.1 any custom factory method (*::Discret()) is loaded from dynamic library
	upon every call to mesher. This was extremely slow and made the mechanism unusable.
	Now library is loaded only once and then its factory method is cached in _DiscretFactory.
	- This requires adding an instance of DataMap<AsciiString,OSD_Function>. To optimize it is added into
	  OSD package and Draw and Plugin packages reuse it.
- Dependency: None


fixes till 370 released as 6.3.1 Update 3
=====================================================


fix371 - ditto fix118.1 (put here to explicitly show that it did not make original 6.3.1 update3).


fix375
- Summary: Initialization of IntPolyh_DynamicArray using object default constructor
- Detailed description: In fact, fix203 assumed that default objects are not used, however this turned to be incorrect.
			IntPolyh_MaillageAffinage::LinkEdges2Triangles() does read values and the algorithm goes wrong
			if the values have not been initialized yet.
- Dependency: fix203
- See also: fix376


fix376
- Summary: Rolling back fix203 and fix375
- Detailed description: In fact, the original fix did not assume object copying during memory extension. However this
	assumption did not prove correct on one example (from cfturbo). The more correct fix should employ use of
	Vector (dynamically growing array which preserves already allocated objects).
	At the moment, the original OCC code is restored thus all performance improvements are rolled back. OCC code
	pre-allocates huge amount of memory (e.g. 30000 objects) while actually using a few hundreds, so pay off is
	tremendous. On the other hand, the code is also not protected against the case when initial allocation is
	insufficient.
- Dependency: None

fixes till 376 released as 6.3.1 Update 4
=====================================================

- Summary:
- Detailed description:
- Dependency: None


- Summary:
- Detailed description:
- Dependency: None