FL_OVERRIDE is defined as `override` for C++11 and higher
FL_OVERRIDE is defined as `override` for VisualC 2015 and newer
Don't interfere with Fl_Widget::override()
- Add internal fl_contrast_cielab() as the new default.
- Keep old function as internal fl_contrast_legacy().
- Add fl_contrast_mode() to switch between fl_contrast() functions.
- Add fl_contrast_level() to fine tune fl_contrast() per mode.
- Add option to register and use a custom contrast function.
- Add test/contrast.cxx test program.
- Move all fl_contrast() related code to a new file src/fl_contrast.cxx.
- Add fl_lightness() convenience function for perceived lightness.
- Add fl_luminance() convenience function for physical luminance.
This commit uses Wayland popup positionning methods to handle common
menu windows and prevents them from expanding below display bottom or above top.
The previous algorithm remains in place for menu windows higher than
the display height. Further changes for these big menus may come later.
- remove static variable 'string_count'
- improve and document methods debug() and hexdump()
This commit does not resolve issue 601 but removes some development
leftovers.
Methods debug() and hexdump() left for debugging as requested.
It may be useful to some to have the terminal emit an error character
to show unknown escape sequences. Off by default, unknown escape sequences
are silently ignored. If enabled, '¿' is inserted instead.
This is basically a rewrite of the ESC handler, keeping state
in the class, so ESC sequences can continued between called
to append() (such as when reading data from a pipe in blocks).
New private class 'Fl_Escape_Seq' handles parsing and state info.
It also has careful bounds checking during parsing.
Backspace supports Unicode, and the unicode chars can straddle
across append() operations as well.
Private variables in Fl_Simple_Terminal renamed _xxx to xxx_
to improve CMP compliance.
SVG is now decompressed in memory
Bitmap invalid array length handling to return an error
RGB Image data reader to return error if image data is too short
FLUID: Add size argument to bitmap and JPEG data
If available, it is used on Linux regardless of the current desktop because
it offers free XDG portal integration, which means it picks the correct file
chooser on all desktops, and allows for meaningful file selection in sandbox
environments like Flatpak.
Change some method names to comply with FLTK style as discussed
in fltk.coredev, thread "Fl_Flex method name question".
* Rename Fl_Flex::margins(...) to Fl_Flex::margin(...)
(use singular form for all margin related methods)
* Remove Fl_Flex::setSize() and isSetSize() "backwards compatibility" methods
* Rename Fl_Flex::set_size(...) to fixed(...)
Note: the latter affects existing (pre-release) fluid (.fl) files: you may
want to edit and replace "set_size_tuples" with "fixed_size_tuples"
Fl_BMP_Image.cxx:228:30: warning: ‘width’ may be used uninitialized in this function
228 | int maskSize = (((width*Bpp+3)&~3)*height) + ...
| ~~~~~^~~~
@wcout reports that we get better results if function pango_font_metrics_get_height()
is not used until version 1.50.6 and above of libpango. The difference is specially visible
with underscore characters in unscaled, standard DPI displays. This commit does that.
[ 15%] Building CXX object src/CMakeFiles/fltk.dir/Fl_Text_Display.cxx.o
.../fltk/src/Fl_Text_Display.cxx: In constructor ‘Fl_Text_Display::Fl_Text_Display(int, int, int, int, const char*)’:
.../fltk/src/Fl_Text_Display.cxx:122:57: warning: ‘void* __builtin_memset(void*, int, long unsigned int)’ writing between 4 and 8589934584 bytes into a region of size 0 overflows the destination [-Wstringop-overflow=]
122 | for (int i=1; i<mNVisibleLines; i++) mLineStarts[i] = -1;
| ~~~~~~~~~~~~~~~^~~~
.../fltk/src/Fl_Text_Display.cxx:120:39: note: at offset 4 into destination object of size 4 allocated by ‘operator new []’
120 | mLineStarts = new int[mNVisibleLines];
| ^
This warning is IMHO obsolete because the code in question should not
be executed at all (mNVisibleLines == 1). However, the compiler seems
to substitute this with '__builtin_memset(...)' and analyzes "correctly"
that memory at offset 4 would be overwritten but not that the written
size would be 0.
The "fix" uses a compiler macro and #if to clarify that this code must
not be executed and should not be compiled (see comment why this code
exists).
