fltk/documentation/osissues.html
Michael R Sweet 87dd7f0d23 Revised documentation files.
git-svn-id: file:///fltk/svn/fltk/trunk@177 ea41ed52-d2ee-0310-a9c1-e6b18d33e121
1998-12-29 14:21:17 +00:00

456 lines
16 KiB
HTML

<HTML>
<BODY>
<H1 ALIGN=RIGHT>F - Operating System Specific Issues</H1
</BODY>
</HTML>
<title>Fltk X-specific interface</title>
<h2>Fltk X-specific interface</h2>
<b>#include &lt;FL/x.H></b>
<p>On Unix/X you can include this file to access fltk's X-specific
functions. Be warned that some of the structures and calls in it are
subject to change in future version of fltk. Try to avoid doing this
so your code is portable.
<p><hr>
<h2>Handling other X events</h2>
<a name=add_handler>
</ul><h4><code>void Fl::add_handler(int (*f)(int));</code></h4><ul>
Install a function to parse unrecognized events. If fltk cannot figure
out what to do with an event, it calls each of these functions (most
recent first) until one of them returns non-zero. If none of them
returns non zero then the event is ignored.
<p>Fltk calls this for any X events it does not recognize, or X events
with a window id that fltk does not recognize. You can look at the X
event with the <a href=#fl_xevent><code>fl_xevent</code></a>.
<p>The argument is zero for unrecognized X events. These handlers are
also called for global shortcuts and some other events that the widget
they were passed to did not handle. In this case the argument is
non-zero (for instance FL_SHORTCUT).
<a name=fl_xevent>
</ul><h4><code>extern XEvent* fl_xvent;</code></h4><ul>
The most current X event.
<a name=fl_event_time>
</ul><h4><code>extern ulong fl_event_time;</code></h4><ul>
This is the time stamp from the most recent X event that reported it
(not all do). Many X calls (like cut and paste) need this value.
</ul><h4><code>Window fl_xid(const Fl_Window*);</code></h4><ul>
Returns the xid for a window, or zero if not shown().
</ul><h4><code>Fl_Window* fl_find(ulong xid);</code></h4><ul>
<p>Return the Fl_Window that corresponds to the given xid, or
null if not found. This uses a cache so it is slightly faster than
iterating through the windows yourself.
</ul><h4><code>int fl_handle(const XEvent&);</code></h4><ul>
This call allows you to supply the X events to fltk, which may allow
fltk to cooperate with another toolkit or library. The return value
is true if fltk understood the event (if the window does not belong to
fltk and the add_handler() functions all ignore it this returns
false).
<p>Besides feeding events your code should call <a
href=Fl.html#flush>Fl::flush()</a> periodically so that fltk redraws
it's windows.
<p>This function will call the callback functions. It will not return
until they complete. In particular if a callback pops up a modal
window (by calling <a href=utilities.html#fl_ask>fl_ask()</a>, for
instance) it will not return until the modal function returns.
</ul>
<p><hr>
<a name=draw>
<h2>Drawing using Xlib</h2>
</ul><h4><code>
extern Display* fl_display;<br>
extern Window fl_window;<br>
extern GC fl_gc;<br>
extern int fl_screen;<br>
extern XVisualInfo* fl_visual;<br>
extern Colormap fl_colormap;</code></h4><ul>
These global variables are set before Fl_Widget::draw() is called, or
by <a href=Fl_Window.html#make_current>Fl_Window::make_current()</a>
You must use them to produce Xlib calls. Don't attempt to change
them. A typical X drawing call is written like this:
<ul><p><code>XDrawSomething(fl_display, fl_window, fl_gc, ...);</code></ul>
<p>Other information such as the position or size of the X window can be
found by looking at <a
href=Fl_Window.html#make_current><code>Fl_Window::current()</code></a>,
which returns a pointer to the Fl_Window being drawn.
</ul><h4><code>unsigned long fl_xpixel(Fl_Color i);</code></h4><ul>
Returns the X pixel number used to draw the given fltk color index.
This is the X pixel that <a href=Draw.html#fl_color>fl_color(i)</a> would use.
</ul><h4><code>unsigned long fl_xpixel(uchar r, uchar g, uchar
b);</code></h4><ul>
Return the X pixel number used to draw the given rgb color. This is
the X pixel that <a href=Draw.html#fl_rgbcolor>fl_color(r,g,b)</a> would use.
</ul><h4><code>extern XFontStruct* fl_xfont;</code></h4><ul>
Points at the font selected by the most recent <a
href=Draw.html#fl_font>fl_font(font,size)</a>. This is not
necessarily the current font of fl_gc (that's not set until <a
href=Draw.html#text>fl_draw(const char*,...)</a> is called).
</ul>
<p><hr>
<a name=visuals>
<h2>Changing the display, screen, or X visual</h2>
<p>Fltk uses only a single display, screen, X visual, and X colormap.
This greatly simplifies it's internal structure and makes it much
smaller and faster. You can change which it uses by setting global
variables <i>before the first Fl_Window::show() is called</i>.
You may also want to call <a
href=Fl.html#visual>Fl::visual(int)</a>, which is a portable interface
to get a full color and/or double buffered visual.
</ul><h4><code>int Fl::display(const char *)</code></h4><ul>
Set which X display to use. This actually does
<code>setenv("DISPLAY",&nbsp;x)</code> so that child programs will
display on the same screen if called with exec(). This must be done
before the display is opened. This call is provided on MSWindows, but
it just sets the environment variable and has no other effect.
</ul><h4><code>extern Display* fl_display;</code></h4><ul>
The open X display. This is needed as an argument to most Xlib calls.
Don't attempt to change it! This is zero before the display is opened.
</ul><h4><code>void fl_open_display();</code></h4><ul>
Open the display. Does nothing if it is already open. This will make
sure <code>fl_display</code> is non-zero. You should call this if you
wish to do X calls and there is a chance that your code will be called
before the first show() of a window.
<p>This may call Fl::abort() if there is an error opening the display.
</ul><h4><code>void fl_close_display();</code></h4><ul>
This closes the X connection. You do <i>not</i> need to call this to
exit, and in fact it is faster to not do so! It may be useful to call
this if you want your program to continue without the X connection.
You cannot open the display again, and probably cannot call any fltk
functions.
</ul><h4><code>extern int fl_screen;</code></h4><ul>
Which screen to use. This is set by fl_open_display() to the default
screen. You can change it by setting this to a different value
immediately afterwards. It can also be set by changing the last
number in the Fl::display() string: host:0,#.
</ul><h4><code>extern XVisualInfo* fl_visual;<br>
extern Colormap fl_colormap;</code></h4><ul>
<p>The visual and colormap that fltk will use for all windows. These
are set by fl_open_display() to the default visual and colormap. You
can change them before calling show() on the first window. Typical
code for changing the default visual:
<ul><code><pre>
Fl::args(argc, argv); // do this first so $DISPLAY is set
fl_open_display();
fl_visual = find_a_good_visual(fl_display, fl_screen);
if (!fl_visual) Fl::abort("No good visual");
fl_colormap = make_a_colormap(fl_display, fl_visual->visual, fl_visual->depth);
// it is now ok to show() windows:
window->show(argc, argv);
</pre></code></ul>
</ul>
<p><hr>
<a name=window>
<h2>Using a subclass of Fl_Window for special X stuff</h2>
Fltk can manage an X window on a different screen, visual and/or
colormap, you just can't use fltk's drawing routines to draw into it.
But you can write your own draw() method that uses Xlib calls only.
<p>Fltk can also manage xid's provided by other libraries or programs,
and call those libraries when the window needs to be redrawn.
<p>To do this, you need to make a subclass of <a
href=Fl_Window.html>Fl_Window</a> and override some of these virtual
functions:
</ul><h4><code>virtual void Fl_Window::show()</code></h4><ul>
<p>If the window is already shown() this must cause it to be raised,
this can usually be done by calling Fl_Window::show(). If not shown()
your implementation must call either Fl_X::set_xid() or
Fl_X::make_xid():
<h4><code>Fl_X* Fl_X::set_xid(Fl_Window*, Window xid);</code></h4><ul>
Allocate a hidden structure called an Fl_X, put the xid into it, and
set a pointer to it from the Fl_Window. This causes
Fl_Window::shown() to return true.
</ul><h4><code>void Fl_X::make_xid(Fl_Window*, XVisualInfo* =
fl_visual, Colormap = fl_colormap);</code></h4><ul>
This static method does the most onerous parts of creating an X window,
including setting the label, resize limitations, etc. It then does
set_xid with this new window and maps the window.
</ul>
<p>Example:
<pre>
void MyWindow::show() {
if (shown()) {Fl_Window::show(); return;} // you must do this!
fl_open_display(); // necessary if this is first window
// we only calcualte the necessary visual & colormap once:
static XVisualInfo *visual;
static Colormap colormap;
if (!visual) {
visual = figure_out_visual();
colormap = XCreateColormap(fl_display, RootWindow(fl_display,fl_screen),
vis->visual, AllocNone);
}
Fl_X::make_xid(this, visual, colormap);
}
</pre>
</ul><h4><code>virtual void Fl_Window::flush()</code></h4><ul>
This virtual function is called by Fl::flush() to update the window.
For fltk's own windows it does this by setting the global variables
fl_window and fl_gc and then calling the draw() method. For your own
windows you might just want to put all the drawing code in here.
<p>The X region that is a combination of all damage() calls done so
far is in <code>Fl_X::i(this)->region</code>. If null then
you should redraw the entire window. The undocumented function
<code>fl_clip_region(XRegion)</code> will initialize the fl clip stack
with a region or null for no clipping. You must set region to null
afterwards as fl_clip_region() now owns it and will delete it when
done.
<p>If <code>damage()==2</code> then only X expose events have
happened. This may be useful if you have an undamaged image (such as
a backing buffer) around.
<p>Here is a sample where an undamaged image is kept somewhere:
<pre>
void MyWindow::flush() {
fl_clip_region(Fl_X::i(this)->region);
Fl_X::i(this)->region = 0;
if (damage() != 2) {... draw things into backing store ...}
... copy backing store to window ...
}
</pre>
</ul><h4><code>virtual void Fl_Window::hide()</code></h4><ul>
Destroy the window server copy of the window. Usually you will
destroy contexts, pixmaps, or other resources used by the window, and
then call Fl_Window::hide() to get rid of the main window identified
by xid(). If you override this, you must also override the destructor
as shown:
<pre>
void MyWindow::hide() {
if (mypixmap) {
XFreePixmap(fl_display,mypixmap);
mypixmap = 0;
}
Fl_Window::hide(); // you must call this
}
</pre>
</ul><h4><code>virtual void Fl_Window::~Fl_Window()</code></h4><ul>
Because of the way C++ works, if you override hide() you <i>must</i>
override the destructor as well (otherwise only the base class hide()
is called):
<pre>
MyWindow::~MyWindow() {
hide();
}
</pre>
</ul>
<p><a href = index.html>(back to contents)</a>
<title>Fltk MSWindows-specific interface</title>
<h2>#include &lt;FL/x.H><br>
Fltk MSWindows-specific interface</h2>
The &lt;FL/x.H> header file defines the interface to fltk's
MSWindows-specific functions. Be warned that some of the structures
and calls in it are subject to change in future version of fltk. Try
to avoid doing this so your code is portable.
<p><hr>
<h2>Handling other MSWindows messages</h2>
<p>A single WNDCLASSEX called "FLTK" is created. All
Fl_Windows are of this class. This window class is created the first
time Fl_Window::show() is called.
<p>You can probably combine fltk with other libraries that make their
own MSWindows window classes. The easiest way is to call Fl::wait(), it
will call DispatchMessage for all messages to the other windows. If
necessary you can let the other library take over (as long as it calls
DispatchMessage()), but you will have to arrange for the function
Fl::flush() to be called regularily (otherwise widgets will not
update), and timeouts and the idle function will not work.
<a name=fl_msg>
</ul><h4><code>extern MSG fl_msg;</code></h4><ul>
The most recent message read by GetMessage (which is called by
<a href=Fl.html#wait>Fl::wait()</a>. This may not be the most recent
message sent to an fltk window, because silly MSWindows calls the
handle procedures directly for some events (sigh).
<a name=add_handler>
</ul><h4><code>void Fl::add_handler(int (*f)(int));</code></h4><ul>
Install a function to parse unrecognized messages sent to fltk
windows. If fltk cannot figure out what to do with a message, it
calls each of these functions (most recent first) until one of them
returns non-zero. The argument passed to the fuctions is zero. If
all the handlers return zero then fltk calls DefWindowProc(...).
</ul><h4><code>HWND fl_xid(const Fl_Window*);</code></h4><ul>
Returns the window handle for a Fl_Window, or zero if not shown().
</ul><h4><code>Fl_Window* fl_find(HWND xid)</code></h4><ul>
<p>Return the Fl_Window that corresponds to the given window handle,
or null if not found. This uses a cache so it is slightly faster than
iterating through the windows yourself.
</ul>
<p><hr>
<a name=gdi>
<h2>Drawing things using the MSWindows GDI</h2>
<p>When the virtual function Fl_Widget::draw() is called, fltk has
stashed in some global variables all the silly extra arguments you
need to make a proper GDI call. These are:
</ul><h4><code>
extern HINSTANCE fl_display;<br>
extern HWND fl_window;<br>
extern HDC fl_gc;<br>
COLORREF fl_RGB();<br>
HPEN fl_pen();<br>
HBRUSH fl_brush();</code></h4><ul>
These global variables are set before draw() is called, or by <a
href=Fl_Window.html#make_current>Fl_Window::make_current()</a> You can
refer to them when needed to produce GDI calls. Don't attempt to
change them. The functions return GDI objects for the current color
set by fl_color(), these are created as needed and cached. A typical
GDI drawing call is written like this:
<ul><p><code>DrawSomething(fl_gc, ..., fl_brush());</code></ul>
<p>It may also be useful to refer to <a
href=Fl_Window.html#make_current><code>Fl_Window::current()</code></a>
to get the window's size or position.
</ul>
<p><hr>
<h2>How to not get a MSDOS console window</h2>
MSWindows has a really stupid mode switch stored in the executables that
controls whether or not to make a console window (hint to Mr Gates:
why not leave it hidden until the program prints something?).
<p>To not produce a "console" window when you run your program add the
following secret incantation to the Micro$oft linker:
<p><pre> /SUBSYSTEM:WINDOWS /ENTRY:mainCRTStartup</pre>
<p>Unfortunately this seems to completely disable stdin/stdout, even
if you run the program from a console. So don't do this until you
have debugged your program!
<p><hr>
<h2>Other hints</h2>
<p>I use capital C as the extension for c++ source code, for instace
for Fluid output. Unfortunately there is no way to convince VC++ to
use this except to tell it to compile *everything* using C++ by
putting the switch "/TP" in the options. This makes it impossible to
combine old C code and fltk code.
<p><hr>
<h2>Known bugs</h2>
<p>If program is deactivated, Fl::wait() does not return until it is
activated again, even though many events are delivered to the program.
This can cause idle background processes to stop unexpectedly. This
also happens while the user is dragging or resizing windows or
otherwise holding the mouse down. I was forced to remove most of the
efficiency fltk uses for redrawing in order to get windows to update
while being moved. This is a design error in MSWindows and probably
impossible to get around.
<p>Fl_Gl_Window::can_do_overlay() returns true until the first time it
attempts to draw an overlay, and then correctly returns whether or not
there is overlay hardware.
<p>Cut text contains ^J rather than ^M^J to break lines. This is a
feature, not a bug.
<p>I can't seem to get SetCapture (used by Fl::grab()) to work, and I
can't find a way to stop the main window title from turning gray while
menus are popped up.
<p>glpuzzle program does not animate unless you resize it first. Unknown why.
<p>Fl_Window::fullscreen() not implemented (should take over the screen
without a title bar). Currently does maximize instead.
<p>Import .bmp files into fluid. Wonko has the specs.
<p>Can't set icon of windows.