We use the host system's `fluid` when cross-compiling. This must be
executable as `fluid`, i.e. it must be in the PATH or otherwise
defined, for instance as an alias.
Add argument 'center' to position the message box centered over
the given x/y coordinates.
Add another method to supply a widget or window to center the
message box over.
Fix documentation and don't use INT_MIN to avoid having to
include limits.h in user code.
Metrowerks CodeWarrior was an ancient macOS compiler (discontinued
since 2005) that defined the macro __MWERKS__. Code using this macro
and several comments have been removed.
That doc says:
"Top-level windows initially have visible() set to 0 and parent() set to NULL"
and here sw is created as a top-level window, so it must be made visible
either by set_visible() or by show().
The demo exhibits a sometimes annoying /feature/ to spin the entire
puzzle if a user happens to click and/or drag the mouse somewhere on
the puzzle's background rather than on a puzzle piece.
A new command line option '-n' disables this feature entirely.
The extended menu and keyboard commands learned a new option and
keyboard shortcut 'n' => "Normal Position" to stop spinning and
to reset the puzzle to its original position.
The space key is another (hidden) shortcut for 'n'.
MacOS uses bundles instead of executables. CMake creates those bundles in various locations, depending on the generator used (Xcode or Makefiles). I tried to fix all instances where demo apps did not find the resources they needed. This probably must be done for Linux and MSWindows as well.
Usage of Fl::readqueue() is not recommended (should be deprecated?),
hence we shouldn't use it in our demo program(s).
To do: remove Fl::readqueue() usage from fluid.
This adds source png and xcf files for checkers pieces.
Also adds Makefile rules using xxd to convert .png -> .h as requested by Albrecht
in issue #29, (hmm, no comment numbers github?)..message dated "June 22nd 2019 4:22am PDT".
OP noticed our one-bit xbm images demonstrated an X11 driver bug
on certain hardware. While it's good that our test programs show the bug
(e.g. test/bitmap), it's bad that our games do.
Removed the 16 xbm files used by checkers which simulated transparency
with dithered images, and replaced with 4 "hexdump" png files that are
cleaner/simpler looking for such lores pieces (46x46 pixels).
Will follow up with a separate commit for the source .png and gimp .xcf
file I used for making the pieces.
The Cairo demo program (test/cairo_test.cxx) can and should be built
even if Cairo is not configured with CMake as it is done with
autotools (configure/make).
The message window shown when Cairo support is not available has been
improved (is more explicit and has a window title).
The intention is to improve (and minimize) diffs when dependencies
are changed and `make depend' is executed.
(1) use `makedepend ... -w 20'
(2) `sort -u' the resultant file with some more "magic".
(1) makes sure that there is only one dependent file per line,
(2) makes lines unique since `makedepend' can output one file
many times which is redundant and makes diffs hard to read.
This uses 'mv', 'grep', and 'sort' in all FLTK library Makefiles.
Image libraries are intentionally left for a later update.
test/fracviewer.cxx wouldn't be compiled if the ABI version was
changed with configure and `make' was executed w/o `make clean'.
Add 'fracviewer.cxx' to list of source files to fix this.
Also add missing unittest_scrollbarsize.cxx to dependencies of
unittests.o.
Finally run `makedepend' to update dependencies.
Note: cherry-picked from branch-1.3 and adjusted to fit master (1.4)
