This commit adds support of another type of composed characters: flags.
It also fixes Fl_Text_Buffer::prev_char() and Fl_Text_Buffer::next_char()
that must use Fl_Text_Buffer::byte_at() to access to the content of the text buffer.
This commit makes platforms Windows and macOS compute string widths
with the same mechanism as what is in place for platforms Wayland/X11:
- the width of a string containing a single codepoint is computed and
memorized in the table of character widths;
- the width of a string containing several codepoints is computed as
such rather than as the sum of the widths of its composing characters.
The result is that FLTK text widgets input and draw correctly also
complex emojis encoded with context-dependent codepoints.
Function fl_utf8_remove_context_dependent() is no longer necessary.
This commit introduces function fl_utf8_remove_context_dependent() that removes
from an UTF-8 string its context-dependent codepoints. Platforms macOS, Wayland
and X11 call this function to process UTF-8 text received from a character palette
as input to FLTK text. This makes sure FLTK text-editing widgets process textual input
equally and consistently across platforms, especially emojis entered via a palette.
Platform Windows creates a series of separate system events to input an emoji
via the character palette. For this reason, function fl_utf8_remove_context_dependent()
is not used by this platform which does internally the same filtering of context-
dependent codepoints.
The character palette allowing to input emojis in text generates in some cases a series
of unicode points to represent a single emoji. These series contain various kinds of
unicode points with context-dependent meaning. This commit prevents such context-
dependent unicodepoints from being inserted in FLTK text because FLTK text edition
mechanism is not ready to handle properly context dependency in edited UTF-8 text.
This fixes input of emojis to Fl_Input and Fl_Text_editor widgets under Windows with the emoji palette.
Most emojis have a Unicode point > 0xFFFF and therefore are encoded as a surrogate pair
by Windows which uses UTF-16. Thus, Windows sends 2 consecutive WM_CHAR messages to the
window and gives one member of the pair each time. After the second WM_CHAR message arrived
FLTK is able to enter the emoji in its text. Windows may also send "variation selectors" and
zero-width Unicode points when dealing with emojis. FLTK just skips them.
Windows also translates some Unicode emojis into 1 emoji + 1 other Unicode point: for example
"woman pilot" produces "pilot emoji" + "woman" unicode point. FLTK now handles
this gracefuly.
This fix also prefixes the windows class names with "FLTK-" under Windows to prevent
collisions with Windows-reserved class names. That fix is necessary for the emoji palette
to be usable in some scenarios. That fix is still under debate and may evolve in latter commits.
This option can be used to disable pen/tablet support if there are
build problems on a particular platform or build system (e.g. MinGW)
so users can continue to build FLTK 1.5.
Users can also choose to disable pen support if they don't need it.
This removes the need to guess names of files each theme gives to cursor shapes
and makes linking with dbus superfluous when the compositor supports
the new protocol.
The old, surface-based approach to cursor shapes remains used for custom shapes.
This rewriting of the FLTK callback function that runs when there are data available
for reading in the socket connecting the app and the Wayland compositor is meant to
facilitate the integration of Vulkan.
This rewriting reproduces the recommended code to read from the socket
documented in Wayland function wl_display_prepare_read_queue() when several threads
potentially read from the socket.
This method can be used to set a more appropriate color average to
prevent "graying out" the box colors of the 'plastic' scheme.
Alternatively environment variable 'FLTK_PLASTIC_AVERAGE' can be used
to set the color average value. See docs for details.
Set color average to 45% in test/unittests demo program.
- The removed code had been disabled in the year 2003 or earlier for
reasons mentioned in those old commits and has never been officially
used again. Use `git blame` to find these commits.
- Update comments and copyright.
- Remove empty lines.
