* Change usage of sample shader code to one with compatible license.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
A loft or pocket only makes sense if the segments used for it have some minimum separation. Having the shapes identical causes segfaults in OCCT
Checking for CoG is a hack. Identical shapes have identical CoG - but an identical CoG does not necessarily mean identical shapes - however it can be argued that a shape that is not offset at least somewhat in the direction of the loft doesnt make a valid shape for a loft (it will typically freeze OCCT instead of crashing it) and a sufficiently suitable shape will have a different CoG
throwing the exception will make it easy to identify if there ever is a case where this is a bad thing - and exceptions are better than segfaults.
This is a stopgap measure, I don't know if the underlying issue is fully understood - but this prevents it from happening in the given test case.
By overriding the base for the pocket prism, which is a trick also applied by
TopoShape::makeElementPrismUntil() if the prism creation fails on first try, the bug where said prism goes to the wrong shape is avoided - this on its own would however break a pad in the inverse case, so we only do that when pocketing
The restored Euler angles of a rotation may significantly differ from the input Euler angles so that determining the plane isn't very reliable.
To get the plane reliably multiply (0,0,1) with the rotation and use this as reference normal of the plane.
Returning a null shape would have triggered an exception in the bottom branch but not in the top one - which later leads to a segfault - solution - raise an exception. OOC doesn't like null shapes.
The measurement code did not handle mixed selections of known and unknown types correctly. Although a unknown selected object would leave the type at Invalid, selecting both known and unknown objects would have the type determined by the known object type and treat the unknown as the same. This causes exceptions and - worst case a segmentation fault.
This fix introduces a new category "other" which - if present in a selection always forces type to Invalid. This should be forward compatible in case new types are introduced in the future.
* Loft Intelligence - Select the whole sketch if the selected shape is a component of a sketch that is not a vertex (Fix#16630)
In 301194a69682633debfc919c4e146811d1b29617 FeatureLoft.cpp was refactured, dropping some functionality, mainly the ability to create lofts to entire sketches or other 2d shapes if a single component that was not a single vertex was selected as either the base shape or a section
the old code was:
1 auto getSectionShape =
2 [](App::DocumentObject* feature, const std::vector<std::string> &subs) -> TopoDS_Shape {
3 if (!feature ||
4 !feature->isDerivedFrom(Part::Feature::getClassTypeId()))
5 throw Base::TypeError("Loft: Invalid profile/section");
6
7 auto subName = subs.empty() ? "" : subs.front();
8
9 // only take the entire shape when we have a sketch selected, but
10 // not a point of the sketch
11 if (feature->isDerivedFrom(Part::Part2DObject::getClassTypeId()) &&
12 subName.compare(0, 6, "Vertex") != 0)
13 return static_cast<Part::Part2DObject*>(feature)->Shape.getValue();
14 else {
15 if(subName.empty())
16 throw Base::ValueError("No valid subelement linked in Part::Feature");
17 return static_cast<Part::Feature*>(feature)->Shape.getShape().getSubShape(subName.c_str());
18 }
19 };
this commit forward-ports the missing functionality provided by line 7-12 in above snippet
* Code cleanup as suggested by (#16791)
* Enforce LC_NUMERIC=C on Posix systems (#16724)
Initialising Qt with native LC_NUMERIC affects XML import on CLI (freecad -c)
This causes sketch constraints and potentially other components relying on numeric constants to break or have incorrect values.
Forcing LC_NUMERIC to C avoids this issue and enforces reproducible behaviour on all locales.
This does not affect the number format displayed in the GUI, as set in preferences.
* Updated ts files
* merged crowdin translations
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* Gui: Minor polishing of cursor regression fix
* TD: Buggy Wayland custom cursors workaround
Until Qt v6.6 displaying custom cursors on Wayland is broken,
so add a workaround. See also QTBUG-95434.
Fix is the same as 094c1b10 ("Gui: Buggy Wayland custom cursors workaround")
plus subsequent fixes.
Recent CMake (>= 3.28) do not allow exec_program() any more by default,
but prefer execute_process(), which takes slightly different flags.
According to the documentation, exec_program() has been deprecated
since CMake 3.0 (which was released in 2014, so it should be common
enough that I don't feel like making allowances for any really old CMake).
References: https://cmake.org/cmake/help/latest/policy/CMP0153.html
* [TD]add method to clean win filespecs
- '\' in strings passed to Python as filespecs is interpreted as
an escape of the following character.
- replace '\' with '/'
* [TD]remove '\' from filespecs before use
* Fix double-import on exception in ImportGUI.readDXF
The python code has a try/catch block intended to detect if the ImportGUI module is present and if so use its readDXF method, otherwise use Import.readDXF.
The block was incorrectly structured so that Import.readDXF would also be called if ImportGUI.readDXF raised an exception, causing the DXF file contents to be loaded twice into the drawing (at least, up to the point where the exception occurred)
* Make ImpExpDxfRead::MakeLayer use centralized accessor for Draft module
The importer class already had a method to find the python Draft module, but MakeLayer was doing the work itself. Now it uses the centralized accessor, which has also been improved to generate a message if the module can't be loaded.
* Give compounded objects names related to the containing layer name
If "Use Layers" is set and also "Group layers into blocks" the names of the generated compound objects will be based on the name of the containing layer.
If "Use Layers" is not set this will not occur because in general objects from several layers would be compounded together.
Fixes (partially) #16068, the remaining fix is likely just an explanation of the interaction of the options.
closes#13597
* Use correct (new) property name OverrideShapeAppearanceChildren
This corrects a bug created by commit 3aea798 which renamed the layer property "OverrideShapeColorChildren" to "OverrideShapeAppearanceChildren" but missed this particular reference to the property by its old name.
The code here called PyObject_SetAttrString which, due to the wrong attribute name, set an expection state in the Python engine, ultimately causing the next attempt at importing a module to fail, with various consequences depending on why the module is being imported.
* Wrap PyObject_SetAttrString and PyObject_GetAttrString with error-checkers
In DEBUG compiles these methods are wrapped to report errors which can occur if the property cannot be referenced.
* Make some error-printers static instead of const
They don't need the CDxfRead object to work
* Display exceptions raised by ReadEntity
Although the DXF reader has an ignore-errors flag, it should still report any errors it encounters. The flag is deemed to mean that, after an error, the reader should continue to try to read the file rather than quitting on the first error.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Unfortunately, not all documentation is translated.
Instead of presenting to the user 404 code error pages is preferable to
show the documentation on English.
Status codes 200 and 404 are assumed as requests response.
Correct the installation paths affecting builds on various systems
while still maintaining the old material editor required for
backwards compatibility
