Also, remove this text from the doc of Fl_Menu_::down_box()
"If this is FL_NO_BOX then it acts like
FL_THIN_UP_BOX and selection_color() acts like
FL_WHITE, for back compatibility."
that was true only for the Windows platform and that required
selection_color to be replaced by white for menu items which is
not what FLTK 1.4 expects.
The new state of menu item drawings is as follows :
- all platforms draw menu items with the same symbolic colors
- selected items and menu titles are drawn with the selection color
as background color
- menu items are drawn by default with no box but can be given one
by Fl_Menu_::down_box(Fl_Boxtype)
- the text of selected items is drawn with the menu's color
unless fl_contrast() finds it does not make enough contrast with
the background (selection color) and substitutes it with a more
adapted color
- the Windows platform uses a visibly different hue for its default
selection color from what other platforms use
... in thread: "FLTK 1.4 Menu Bar Style", started on May 6, 2024
This is about the little triangle (or similar) pointing right when
a submenu exists in a menu. The arrow form differs per scheme.
... as reported in thread "Windows 11 crash (still investigating)"
of fltk.coredev, started on Apr 24, 2024.
This bug could only be triggered if a /subwindow/ wss hidden but was
not limited to Windows.
The commit f288aea from Feb. 23rd introduced a
regression where children of subwindows were moved
inside the subwindow when only the subwindow itself
was supposed to move.
- remove build option FLTK_USE_KDIALOG replaced by an Fl::option() option
- new run-time option OPTION_FNFC_USES_KDIALOG
- make options OPTION_FNFC_USES_KDIALOG and OPTION_FNFC_USES_ZENITY
false by default
- add mention of new program fltk-options in the doc of Fl::option()
- change logic of choice of the native file chooser under X11/Wayland:
the zenity and kdialog choosers are opt-in; otherwise the GTK file chooser
is used, unless opted out with OPTION_FNFC_USES_GTK
- document that zenity may be interesting for sandboxed apps
- document that both zenity and kdialog make member functions
Fl_Native_File_Chooser::filter_value() inoperable
This ensures that header files for libraries found by `pkg-config`
(pkg_check_modules) in non-system directories will be found by the
compiler during the compilation of the FLTK library.
This issue has been reported in PR #954 related to builds with Conan
where some header files are not in system locations.
Unfortunately commit 5417ea5f1f broke
simple user projects by propagating unknown CMake target names to
linker requirements of user projects.
This commit tries to fix this w/o breaking the intentions of PR #954.
More precise: ... if the last picked item is in a "detached submenu",
i.e. in one addressed by another menu item with the FL_SUBMENU_POINTER
flag set.
Also: document *why* this is the case, and what the result value is.
* CMake integration, no autotiools
* alignment panel is now correctly renamed to setting panel
* source view is now correctly renamed to code view
* Merge FLTK FLUID docs into FLUID user manual.
* Add two simple entry tutorials
* Remove FLUID chapter form FLTK docs.
* GitHub action to generate HTML and PDF docs and
make the available as artefacts
Parts of this commit fix the usage of CMake targets in the generation
process of fltk-config modified by PR #954.
The rest improves handling of variables used to generate fltk-config,
removes obsolete comments, documents variables like GLLIBS etc. which
had to be done anyway.
This carries include directories and the required link flags.
Several libraries brought in through pkg_check_modules aren't linked correctly using the _LDFLAGS variables.
Instead, they link the libraries directly with `-l`.
This is problematic because it doesn't properly pull in the actual directory of the library being linked when it is not in a system library directory.
Likewise, the necessary include directories for several of these targets aren't properly set either.
Linking against the imported targets automatically pulls in the necessary include directories.
* CMake: Improve detection of the GLU library and GL/glu.h header file
Locate the GLU library and header independent of the GL library and header locations.
Add the GLU header location to necessary target_include_directory calls.
* CMake: Locate and use the GL include directory
Find the GL include directory and use it in target_include_directories calls.
- Add 'int Fl_Terminal::handle_unknown_char(int drow, int dcol)'
to write the "unknown" character to the intended display position.
- Define Fl_Terminal::unknown_char as a static variable to avoid
redundancy. In the future this might be overridden by users.
Some old platforms (NetBSD, AIX) implement the common stdio "putchar()"
function as a global macro which poisons the global namespace, preventing
all C and C++ programs from using "putchar()" as a function or method.
There was a long thread about this in fltk.coredev during the period
Mar 25 2024 ~ Apr 4 2024, subject "RFC: Fl_Terminal::putchar() in public API"
as to why we have no choice but to not use putchar() as a method name.
libdecor/build/Makefile: unused anyway (previously used building demos)
src/CMakeLists.txt: broke some user builds, see fltk.general, thread
"-no-pie in FLTK Libraries build." (Mar 17, 2024).
Todo: check usage of "-fPIC" (libdecor builds)
Fix usage of vsscanf in Fl_Unix_System_Driver::clocale_vsscanf().
Final commit of all patches provided by the OP of issue #937 - if
everything works now.
Some of these accessor methods should be private so they can't be used
by user code but - due to compiler issues - they must be public for
HP-UX 11.11 (for details see GitHub Issue #937).
Fl_X11_Screen_Driver::set_spot() did not free the list of missing
fonts returned by XCreateFontSet().
This commit also initializes some variables and reformats code, but
the only notable change is to call XFreeStringList(missing_list) if
missing_list is not NULL.
Works now much better with old C99 and C++98 standard compilers.
Fixed: C++ comments in C files and headers included by C files.
There are still some warnings with C90 though but these would be
hard to fix and left as-is for now.
test/fractals.cxx: some arrays were too small by 1, or the compiler
warned at least (false positive?). Anyway, it's fixed now.
- it's no longer necessary to take care of the change in the layout of "struct libdecor"
between versions ≤ 0.2.2 and > 0.2.2 of libdecor
- version > 0.2.2 contains MR131, that is, it defines LIBDECOR_WINDOW_STATE_RESIZING
so the hack to emulate it is no longer necessary
- CMake option FLTK_USE_SYSTEM_LIBDECOR now requires libdecor version > 0.2.2
to be activated, otherwise the bundled libdecor is used
- what will be the libdecor version after 0.2.2 (0.2.3? 0.3.0?) is not known as of today
Thanks to Mike Qin for finding how to query the possibly changed DPI value
of an X11 display while an FLTK app keeps running without running
an external command in a pipe.
This involved enlarging the unittest main window to make room
for the additional test.
Adding this test revealed a problem in Fl_Terminal's global
scrollbar size handling, which is fixed here as well.
Also fixed a small issue in the demo's debugging terminal
with the horiz scrollbar.
The macOS Window menu is connected to the item array of the macOS menubar
via FL_SUBMENU_POINTER. This facilitates memorization of pointers
to items of the macOS menubar because the item array no longer changes
each time a window opens of gets closed.
* Improve horizontal interactive selection
* Using half-character positions to implement selection
similar to Fl_Input.
* Add word and line selection
* Fix vertical position of text
Fixed a bug in the handling of xterm color 39 + 49, which avoids
applying Dim/Bold if default color is the special "see through" color
0xffffff00. A test will be added to test/terminal to excercise this
in a follow up commit.
Renamed CharStyle::?gcolor_uchar() -> CharStyle::?gcolor_xterm()
for a consistent naming convention.
Changed fltk_fg_color() from a static func to a CharStyle method
because it needs to access the defaultcolor for the special cases
for xterm colors 39 and 49.
Made CharStyle::attr_color() private, as per Jonathan Griffitts request
in issue #909.
Improved docs regarding effects of Dim/Bold for xterm and non-xterm color
methods.
Fl_Terminal::text?gcolor_xterm() methods now use CharStyle::?gcolor_xterm()
methods, removing the duplicate code logic.
This is a big commit and there are too many changes to list them all.
The main changes are:
- rename all CMake build options to 'FLTK_*'
- export library targets with namespace (prefix) 'fltk::'
- standardize shared library target names with suffix '-shared'
- set public build properties on libraries for consumers
- document library names and aliases in README.CMake.txt
- document changes in "Migrating Code from FLTK 1.3 to 1.4"
- partial backwards compatibility for old user projects
Included but not directly related changes:
- fix Windows (Visual Studio) DLL build
- add CMake function fl_debug_target() to show target properties
- don't build test programs if FLTK is a subproject
- internal: reformat CMake code: remove space before '('
Thanks to Matthias and Manolo for their help, testing, and feeback.
Fl_Window_Driver::set_popup_window() is to be used to declare a window should be
positioned relatively to a previously mapped other window. This allows a platform
to process such windows differently from other windows if needed.
Menu and tooltip windows are so declared.
A call to Fl_Window_Driver::set_popup_window() also allows to distinguish a real
menu or tooltip window from a window marked by Fl_Window::set_menu_window()
or by Fl_Window::set_tooltip_window() but that's not a real menu or tooltip.
New member function bool Fl_Screen_Driver::screen_boundaries_known() returns
true by default. A platform where the position of windows inside a screen is hidden
(e.g., Wayland) returns false. This allows FLTK to refrain from trying to make sure
a computed position is inside a screen.
When an Fl_Text_Editor ended up under the mouse, it took
mousewheel events for itself, freezing navigation of a
parent's Fl_Scroll.
Patch handles mousewheel events for both up/down and left/right
by checking:
> If editor has no scrollbar (e.g. no content to scroll to),
then mousewheel events for that scrollbar are ignored
so the parent can be navigated.
> If editor has a scrollbar, it handles wheel events until scroller reaches
an extreme position (e.g. top or bottom), it starts ignoring wheel events
so the parent can be navigated.
Fixes also a similar crash occuring after use of set_tooltip_window().
Two bool member variables are added to class Fl_Window_Driver
to indicate true FLTK menu windows (i.e., created by Fl_Menu_Item::pulldown())
and true FLTK tooltip windows (i.e., Fl_Tooltip).
Member cb of struct wld_buffer is replaced by member frame_cb of struct wld_window.
This allows frame_cb to be used both for non-GL and for top-level GL windows.
Keyboard behavior for button types was not consistent with mouse
click behavior. This change avoids callbacks when pressed radio buttons
are pressed again.
Previous commits introduced too many changes to Fl_Button callbacks,
creating problems in user apps. This is the base for a second much
more specific fix to event flags.
Implement a less strict handling of resizing of a decorated window entirely covered
by a subwindow where throttling of fast resize commands is guided by the
value of the in_use member variable of the window's wld_buffer.
- this commit does rewrite some of the logic,
hoping to make thing more, um, logical. The bits
react mostly the same and hopefully as expected.
Everything is documented in the Fl_Button ctor.
- if a tile with a size range is resize smaller than the
minimum size of all children, the children will now no
longer shrink, and expanding will make them work
again as expected.
Mutter implements too strictly this rule expected from compositors about "frame callbacks" :
"A server should avoid signaling the frame callbacks if the surface is not visible in any way,
e.g. the surface is off-screen, or completely obscured by other opaque surfaces."
When a window is being interactively resized, it makes no sense to create a frame callback
for an entirely covered surface but then never signal the surface can be redrawn,
because this blocks the resize operation.
Neither KWin nor Weston have this problem.
This addresses some issues pointed out by GitHub Issue #875.
Documentation lacked details about Fl_Multi_Label assignment and
correct memory handling.
The new methods
- Fl_Menu_Item::image_label(const Fl_Image *) and
- Fl_Menu_Item::multi_label(const Fl_Multi_Label *)
provide a cleaner interface to assign images and Fl_Multi_Label's
to menu items.
examples/howto-menu-with-images.cxx: carify some issues, fix leak,
and use new Fl_Menu_Item::multi_label(const Fl_Multi_Label *).
This issue was revealed during testing for GitHub Issue #875.
"ERROR: AddressSanitizer: attempting free on address which was not
malloc()-ed", reported by examples/howto-menu-with-images.cxx if
the window object was released at the end of the program, causing
Fl_Menu_::clear() to be called.
The issue was caused by casting all supported label types to
'const char *' which are stored in Fl_Menu_Item::text and then trying
to free() all text strings in Fl_Menu_::clear() under certain
conditions.
Now images and Fl_Multi_Label's are no longer (tried to be) free'd.
This library is no longer needed in FLTK itself. There is only one
demo program (test/forms.cxx) that uses it if it is built. The demo
program displays a message if fltk_forms is not available.
The default is 'enabled' for backwards compatibility but this may
be changed to 'disabled' in a future version.
- the current UI for fallback/default/original shortcut was
confusing, so I remove it until we find something better. I
kept the values so we can reinstate a better interface without
changing the API.
