with perimeters:
1) Increased accuracy of the contour length parametrization from
float to double, as double should capture the difference of
32bit coord_t with full accuracy (or at least very close).
2) The algorithm to insert the T-joint points into the infill perimeter
contour was improved to avoid inserting duplicate points.
Relies to:
Concentric Fill Start Point - New Feature Request #4948
Feature Request: Archimedean Chords - Option to define direction of travel (Inside-Out or Outside-In) #5214
".gcode", ".gco", ".g" and ".ngc" were considered to be G-code file
extensions by the G-code import / export file dialogs, but not by
various other functions. Now the G-code extension is tested by
a single function is_gcode_file(string).
From the Win 2004 preset combobox lose a focus after change the preset selection and that is why the up/down arrow doesn't work properly.
So, set the focus to the combobox explicitly.
The problem is that PrintObject support generator is only run when m_layers.size() > 1, so one-layer object will skip it.
It apparently never worked, after recent refactoring (probably 8bfc986) it started to crash in GCode generator.
This commit fixes the crash, but not the problem. The raft is still not generated (like in all versions since 2.0.0 at least).
Previous implementation did not use monospace font on Linux. According to a note
from @wavexx in ba7f39a: "wxSYS_ANSI_FIXED_FONT use the same as DEFAULT in wxGtk".
This commit uses the same workaround by actually loading the font from GUI_App.
The issue was handling of the special "default_filament_profile" vector,
which is not being normalized in length to the number of extruders,
as this vector is being shown to the user at the Printer dependencies page,
and we don't want to present empty fields there, if the default filament
profile was not defined in the system preset.
The solution adds a status flag to GUI_App::init_openg and ensures that GUI_App::post_init isn't called before initialization in GUI_App::init_openg is done.
Mark Ejection as failed if the wait() function fails. It seems that it is not simply possible
to retrieve the exit code of the process in that case (although it usually finishes sucessfully).
Problem: Add an object, name it using cyrilic, Slice, click Export: the proposed name is mangled. Affects all platforms, 2.2.0 was fine.
Cause: It is mangled during ASCII folding, which was broken with 28c0880 when it was generalized to be used from the fuzzy matcher.
fold_to_ascii(wchar_t c, char *out) relies on narrowing char = wchar_t assignment, the old one used std::back_insert_iterator<std::wstring>.
It is thus unable to leave a character alone even when it should (it should, right?).
Solution:
1. Typed the fold_to_ascii function so it works on wchar_t only, which should mimic the old behaviour.
2. Changed the respective call in fts_fuzzy_match.h. That function also works with wide char C-strings.
Cleanup:
1. Removed the unused fold_utf8_to_ascii(const char *src) overload to avoid code duplication.
2. Untemplated the fold_to_ascii(wchar_t c, std::back_insert_iterator<std::wstring>& out) function, it was never called with a different type.
3. The function is now static in ASCIIFolding.cpp, nobody else needs to know.
The issue is that opening the wxMessageDialog in GUI_App::check_printer_presets fires wxEVT_IDLE event (on Linux at least).
At that point it is already bound to our lambda which will in turn call post_init prematurely.
The solution is to move the Bind at the very end of the initialization. The post_init that it calls assumes that init has been finished anyway.
Fix some char *foo = "string literal" assignments that
really should be assigned to const char*.
(they also happen to be constexpr, but that only prevents
to assign something later to foo, but does not alter
the char* type).
Signed-off-by: Henner Zeller <h.zeller@acm.org>
