[RFA] libgui/src/tkWarpPointer.c for cygwin

[RFA] libgui/src/tkWarpPointer.c for cygwin

[Keith Seitz]

Hi,

I am working on making the Insight testsuite work (better?), and I wanted
to make sure (as much as possible) that we are able to run the same tests
on Cygwin and unix.

For some tests, it is necessary to warp the pointer. This patch simply
implements this functionality on Windows, too.

Keith

src/libgui/ChangeLog:
2001-05-01  Keith Seitz  <keiths@cygnus.com>

	* src/tkWarpPointer.c (WarpPointer): Implement for windows.

Patch:
Index: src/tkWarpPointer.c
===================================================================
RCS file: /cvs/cvsfiles/devo/libgui/src/tkWarpPointer.c,v
retrieving revision 1.2
diff -p -r1.2 tkWarpPointer.c
*** tkWarpPointer.c	1999/01/28 01:18:27	1.2
--- tkWarpPointer.c	2001/05/02 20:11:06
***************
*** 24,32 ****
  *
  *
  * ---------------------------------------------------------------------------*/
- #ifndef _WIN32
-
  #include "tk.h"

  int
  WarpPointer (clientData, interp, objc, objv)
--- 24,34 ----
  *
  *
  * ---------------------------------------------------------------------------*/
  #include "tk.h"
+ #ifdef _WIN32
+ #include <windows.h>
+ #include <winuser.h>
+ #endif

  int
  WarpPointer (clientData, interp, objc, objv)
*************** WarpPointer (clientData, interp, objc, o
*** 36,47 ****
      Tcl_Obj *CONST objv[];
  {
      Tk_Window tkwin;
-     Window win;
      int x, y;
      char *str;

      if (objc != 4) {
!       Tcl_WrongNumArgs(interp, 1, objv, "x y widgetId");
        return TCL_ERROR;
      }

--- 38,48 ----
      Tcl_Obj *CONST objv[];
  {
      Tk_Window tkwin;
      int x, y;
      char *str;

      if (objc != 4) {
!       Tcl_WrongNumArgs(interp, 1, objv, "widgetId x y");
        return TCL_ERROR;
      }

*************** WarpPointer (clientData, interp, objc, o
*** 53,61 ****
  			 Tk_MainWindow (interp));
      if (tkwin == NULL)
        return TCL_ERROR;

-     win = Tk_WindowId(tkwin);
-     XWarpPointer(Tk_Display(tkwin), None, win, 0, 0, 0, 0, x, y);
      return TCL_OK;
  }

--- 54,71 ----
  			 Tk_MainWindow (interp));
      if (tkwin == NULL)
        return TCL_ERROR;
+
+     {
+ #ifdef _WIN32
+       int wx, wy;
+       Tk_GetRootCoords (tkwin, &wx, &wy);
+       SetCursorPos@8 (wx + x, wy + y);
+ #else
+       Window win = Tk_WindowId(tkwin);
+       XWarpPointer(Tk_Display(tkwin), None, win, 0, 0, 0, 0, x, y);
+ #endif
+     }

      return TCL_OK;
  }

*************** cyg_create_warp_pointer_command (Tcl_Int
*** 66,70 ****
      return TCL_ERROR;
    return TCL_OK;
  }
-
- #endif /* !_WIN32 */
--- 76,78 ----

Re: [RFA] libgui/src/tkWarpPointer.c for cygwin

[Syd Polk]

At 01:12 PM 5/2/01 -0700, Keith Seitz wrote:
>Hi,
>
>I am working on making the Insight testsuite work (better?), and I wanted
>to make sure (as much as possible) that we are able to run the same tests
>on Cygwin and unix.
>
>For some tests, it is necessary to warp the pointer. This patch simply
>implements this functionality on Windows, too.
>
>Keith
>
>src/libgui/ChangeLog:
>2001-05-01  Keith Seitz  <keiths@cygnus.com>
>
>         * src/tkWarpPointer.c (WarpPointer): Implement for windows.
>
>Patch:
>Index: src/tkWarpPointer.c
>===================================================================
>RCS file: /cvs/cvsfiles/devo/libgui/src/tkWarpPointer.c,v
>retrieving revision 1.2
>diff -p -r1.2 tkWarpPointer.c
>*** tkWarpPointer.c     1999/01/28 01:18:27     1.2
>--- tkWarpPointer.c     2001/05/02 20:11:06
>***************
>*** 24,32 ****
>   *
>   *
>   * 
> ---------------------------------------------------------------------------*/
>- #ifndef _WIN32
>-
>   #include "tk.h"
>
>   int
>   WarpPointer (clientData, interp, objc, objv)
>--- 24,34 ----
>   *
>   *
>   * 
> ---------------------------------------------------------------------------*/
>   #include "tk.h"
>+ #ifdef _WIN32
>+ #include <windows.h>
>+ #include <winuser.h>
>+ #endif
>
>   int
>   WarpPointer (clientData, interp, objc, objv)
>*************** WarpPointer (clientData, interp, objc, o
>*** 36,47 ****
>       Tcl_Obj *CONST objv[];
>   {
>       Tk_Window tkwin;
>-     Window win;
>       int x, y;
>       char *str;
>
>       if (objc != 4) {
>!       Tcl_WrongNumArgs(interp, 1, objv, "x y widgetId");
>         return TCL_ERROR;
>       }
>
>--- 38,48 ----
>       Tcl_Obj *CONST objv[];
>   {
>       Tk_Window tkwin;
>       int x, y;
>       char *str;
>
>       if (objc != 4) {
>!       Tcl_WrongNumArgs(interp, 1, objv, "widgetId x y");
>         return TCL_ERROR;
>       }
>
>*************** WarpPointer (clientData, interp, objc, o
>*** 53,61 ****
>                         Tk_MainWindow (interp));
>       if (tkwin == NULL)
>         return TCL_ERROR;
>
>-     win = Tk_WindowId(tkwin);
>-     XWarpPointer(Tk_Display(tkwin), None, win, 0, 0, 0, 0, x, y);
>       return TCL_OK;
>   }
>
>--- 54,71 ----
>                         Tk_MainWindow (interp));
>       if (tkwin == NULL)
>         return TCL_ERROR;
>+
>+     {
>+ #ifdef _WIN32
>+       int wx, wy;
>+       Tk_GetRootCoords (tkwin, &wx, &wy);
>+       SetCursorPos@8 (wx + x, wy + y);
>+ #else
>+       Window win = Tk_WindowId(tkwin);
>+       XWarpPointer(Tk_Display(tkwin), None, win, 0, 0, 0, 0, x, y);
>+ #endif
>+     }

I hate #ifdef's like this, I would prefer there to be two files, one for 
windows and one for unix, with an implementation for either.

However, assuming that nothing else in libgui is implemented that way, I 
guess this is ok.

The second question I have is why do you need "SetCursorPos@8"? Assuming 
you are including winuser.h, you should be able to use "SetCursorPos", and 
either a macro will fix this, or the linker will. I don't think requiring 
the "@8" or "@16" or whatever for Windows API calls is very nice.


>       return TCL_OK;
>   }
>
>*************** cyg_create_warp_pointer_command (Tcl_Int
>*** 66,70 ****
>       return TCL_ERROR;
>     return TCL_OK;
>   }
>-
>- #endif /* !_WIN32 */
>--- 76,78 ----

Syd Polk		spolk@redhat.com
Engineering Manager	+1 408 543 9430
Red Hat, Inc.

Re: [RFA] libgui/src/tkWarpPointer.c for cygwin

[Christopher Faylor]

On Wed, May 02, 2001 at 05:06:10PM -0700, Syd Polk wrote:
>The second question I have is why do you need "SetCursorPos@8"? Assuming 
>you are including winuser.h, you should be able to use "SetCursorPos", and 
>either a macro will fix this, or the linker will. I don't think requiring 
>the "@8" or "@16" or whatever for Windows API calls is very nice.

I agree.  The function should be prototyped.  I believe that it should be
pulled in by  a #include <windows.h> .

cgf

Re: [RFA] libgui/src/tkWarpPointer.c for cygwin

[Keith Seitz]

On Wed, 2 May 2001, Syd Polk wrote:

> >+ #ifdef _WIN32
> >+       int wx, wy;
> >+       Tk_GetRootCoords (tkwin, &wx, &wy);
> >+       SetCursorPos@8 (wx + x, wy + y);
> >+ #else
> >+       Window win = Tk_WindowId(tkwin);
> >+       XWarpPointer(Tk_Display(tkwin), None, win, 0, 0, 0, 0, x, y);
> >+ #endif
> >+     }
>
> The second question I have is why do you need "SetCursorPos@8"? Assuming
> you are including winuser.h, you should be able to use "SetCursorPos", and
> either a macro will fix this, or the linker will. I don't think requiring
> the "@8" or "@16" or whatever for Windows API calls is very nice.

I don't need "SetCursorPos@8": I added the header files and forgot to
rename the function in the patch I posted. The actual patch would have
contained a call to SetCursorPos only. Sorry about that.

Keith

Re: [RFA] libgui/src/tkWarpPointer.c for cygwin

[Syd Polk]

At 07:02 AM 5/3/01 -0700, Keith Seitz wrote:
>On Wed, 2 May 2001, Syd Polk wrote:
>
> > >+ #ifdef _WIN32
> > >+       int wx, wy;
> > >+       Tk_GetRootCoords (tkwin, &wx, &wy);
> > >+       SetCursorPos@8 (wx + x, wy + y);
> > >+ #else
> > >+       Window win = Tk_WindowId(tkwin);
> > >+       XWarpPointer(Tk_Display(tkwin), None, win, 0, 0, 0, 0, x, y);
> > >+ #endif
> > >+     }
> >
> > The second question I have is why do you need "SetCursorPos@8"? Assuming
> > you are including winuser.h, you should be able to use "SetCursorPos", and
> > either a macro will fix this, or the linker will. I don't think requiring
> > the "@8" or "@16" or whatever for Windows API calls is very nice.
>
>I don't need "SetCursorPos@8": I added the header files and forgot to
>rename the function in the patch I posted. The actual patch would have
>contained a call to SetCursorPos only. Sorry about that.
>
>Keith

Can you change it and repost the patch?


Syd Polk		spolk@redhat.com
Engineering Manager	+1 408 543 9430
Red Hat, Inc.

Re: [RFA] libgui/src/tkWarpPointer.c for cygwin

[Keith Seitz]

On Thu, 3 May 2001, Syd Polk wrote:

> Can you change it and repost the patch?

Here it is.
Keith

Index: libgui/src/tkWarpPointer.c
===================================================================
RCS file: /cvs/src/src/libgui/src/tkWarpPointer.c,v
retrieving revision 1.1.1.1
diff -u -r1.1.1.1 tkWarpPointer.c
--- libgui/src/tkWarpPointer.c	2000/02/07 00:19:47	1.1.1.1
+++ libgui/src/tkWarpPointer.c	2001/05/03 17:51:59
@@ -24,9 +24,11 @@
 *
 *
 * ---------------------------------------------------------------------------*/
-#ifndef _WIN32
-
 #include "tk.h"
+#ifdef _WIN32
+#include <windows.h>
+#include <winuser.h>
+#endif

 int
 WarpPointer (clientData, interp, objc, objv)
@@ -36,12 +38,11 @@
     Tcl_Obj *CONST objv[];
 {
     Tk_Window tkwin;
-    Window win;
     int x, y;
     char *str;

     if (objc != 4) {
-      Tcl_WrongNumArgs(interp, 1, objv, "x y widgetId");
+      Tcl_WrongNumArgs(interp, 1, objv, "widgetId x y");
       return TCL_ERROR;
     }

@@ -53,9 +54,18 @@
 			 Tk_MainWindow (interp));
     if (tkwin == NULL)
       return TCL_ERROR;
+
+    {
+#ifdef _WIN32
+      int wx, wy;
+      Tk_GetRootCoords (tkwin, &wx, &wy);
+      SetCursorPos (wx + x, wy + y);
+#else
+      Window win = Tk_WindowId(tkwin);
+      XWarpPointer(Tk_Display(tkwin), None, win, 0, 0, 0, 0, x, y);
+#endif
+    }

-    win = Tk_WindowId(tkwin);
-    XWarpPointer(Tk_Display(tkwin), None, win, 0, 0, 0, 0, x, y);
     return TCL_OK;
 }

@@ -66,5 +76,3 @@
     return TCL_ERROR;
   return TCL_OK;
 }
-
-#endif /* !_WIN32 */

Re: [RFA] libgui/src/tkWarpPointer.c for cygwin

[Syd Polk]

Approved.

At 10:52 AM 5/3/01 -0700, Keith Seitz wrote:
>On Thu, 3 May 2001, Syd Polk wrote:
>
> > Can you change it and repost the patch?
>
>Here it is.
>Keith
>
>Index: libgui/src/tkWarpPointer.c
>===================================================================
>RCS file: /cvs/src/src/libgui/src/tkWarpPointer.c,v
>retrieving revision 1.1.1.1
>diff -u -r1.1.1.1 tkWarpPointer.c
>--- libgui/src/tkWarpPointer.c  2000/02/07 00:19:47     1.1.1.1
>+++ libgui/src/tkWarpPointer.c  2001/05/03 17:51:59
>@@ -24,9 +24,11 @@
>  *
>  *
>  * 
> ---------------------------------------------------------------------------*/
>-#ifndef _WIN32
>-
>  #include "tk.h"
>+#ifdef _WIN32
>+#include <windows.h>
>+#include <winuser.h>
>+#endif
>
>  int
>  WarpPointer (clientData, interp, objc, objv)
>@@ -36,12 +38,11 @@
>      Tcl_Obj *CONST objv[];
>  {
>      Tk_Window tkwin;
>-    Window win;
>      int x, y;
>      char *str;
>
>      if (objc != 4) {
>-      Tcl_WrongNumArgs(interp, 1, objv, "x y widgetId");
>+      Tcl_WrongNumArgs(interp, 1, objv, "widgetId x y");
>        return TCL_ERROR;
>      }
>
>@@ -53,9 +54,18 @@
>                         Tk_MainWindow (interp));
>      if (tkwin == NULL)
>        return TCL_ERROR;
>+
>+    {
>+#ifdef _WIN32
>+      int wx, wy;
>+      Tk_GetRootCoords (tkwin, &wx, &wy);
>+      SetCursorPos (wx + x, wy + y);
>+#else
>+      Window win = Tk_WindowId(tkwin);
>+      XWarpPointer(Tk_Display(tkwin), None, win, 0, 0, 0, 0, x, y);
>+#endif
>+    }
>
>-    win = Tk_WindowId(tkwin);
>-    XWarpPointer(Tk_Display(tkwin), None, win, 0, 0, 0, 0, x, y);
>      return TCL_OK;
>  }
>
>@@ -66,5 +76,3 @@
>      return TCL_ERROR;
>    return TCL_OK;
>  }
>-
>-#endif /* !_WIN32 */

Syd Polk		spolk@redhat.com
Engineering Manager	+1 408 543 9430
Red Hat, Inc.

Insight port to the 'original' Windows

[Kai Ruottu]

Hi,
after a year or so pause, I tried the Win32s version of Insight again,
with the
5.0 + sources (the current snapshots would most probably behave just the
same way....)

The result is attached as an screenshot showing the problem....

 Has someone to give a clue what could cause this?  Is there some
fundamental feature (threads etc.) taken into work into the Insight
scripts etc. in 1999 causing this?  Ok, as seen in the source window,
all the stuff in the lower part has not been drawn....

 The scripts etc. taken from the GNUPro-98r2 sources (those from
Hitachi) work quite well under Win32s, but the GUI from that time
is a little 'clumsy', so I would like to upgrade the base in the
Mingw/Win3.1x hosted Insights.... And after this to put the diffs
somewhere (also for the working Mingw/Win32 one...)

 BTW, I will change the 'gdbtcl/images2/insight.gif' to the original
when I will find it somewhere.... Sometimes the snapshots erased
all the gdbtcl-sources and I see only a red window with a white
background old image with the current sources.... Perhaps the
'insight+dejagnu-*.diff.bz2's never included the current new
'insight.gif'....

 Many thanks in advance !

Cheers, Kai

PS. the false colors came when moving the stuff via clipboard into
PaintShop, in reality the colors are just ok in the toolbar an in
the 'About Window'...

Re: Insight port to the 'original' Windows

[Keith Seitz]

On Thu, 4 May 2000, Kai Ruottu wrote:

> The result is attached as an screenshot showing the problem....

You ARE adventurous, aren't you? :-)

>  Has someone to give a clue what could cause this?  Is there some
> fundamental feature (threads etc.) taken into work into the Insight
> scripts etc. in 1999 causing this?  Ok, as seen in the source window,
> all the stuff in the lower part has not been drawn....

For some reason, the srctextwin has not drawn itself. It could be a bug in
Insight or a bug in Tcl. I would start Insight withe environment varialble
GDBTK_DEBUG set (to two) in your environment. Of course, I don't know if
such a thing exists on the original version of windows, so you may have to
edit your gdbtk.ini. There is a section labeled "windows" with the
property "active={list of windows to open when started". Add "DebugWin".
This will automatically bring up the DebugWindow when you start Insight.
Maybe that will provide some insight into your problem. (Ok, I couildn't
resist that one. :-)

>  The scripts etc. taken from the GNUPro-98r2 sources (those from
> Hitachi) work quite well under Win32s, but the GUI from that time
> is a little 'clumsy', so I would like to upgrade the base in the
> Mingw/Win3.1x hosted Insights.... And after this to put the diffs
> somewhere (also for the working Mingw/Win32 one...)

Unfortunately, a whole lot has changed since then. There is almost no way
of knowing what broke it.

Let us know if the DebugWin shows any errors...
Keith

Re: Insight port to the 'original' Windows

[Kai Ruottu]

Keith Seitz wrote:

> On Thu, 4 May 2000, Kai Ruottu wrote:
>
> > The result is attached as an screenshot showing the problem....
>
> You ARE adventurous, aren't you? :-)
>

 Perhaps....

> >  Has someone to give a clue what could cause this?  Is there some
> > fundamental feature (threads etc.) taken into work into the Insight
> > scripts etc. in 1999 causing this?  Ok, as seen in the source window,
> > all the stuff in the lower part has not been drawn....
>
> For some reason, the srctextwin has not drawn itself. It could be a bug in
> Insight or a bug in Tcl. I would start Insight withe environment varialble
> GDBTK_DEBUG set (to two) in your environment. Of course, I don't know if
> such a thing exists on the original version of windows, so you may have to
> edit your gdbtk.ini. There is a section labeled "windows" with the
> property "active={list of windows to open when started". Add "DebugWin".
> This will automatically bring up the DebugWindow when you start Insight.
> Maybe that will provide some insight into your problem. (Ok, I couildn't
> resist that one. :-)

 Unfortunately the Debug Window is one of those windows not starting :-(
The others not starting are the Memory, Console and HelpTopics while all
kind of others seem to work at least somehow....  Setting the GDBTK_DEBUG
adds the Debug Window -choice into the View-pulldown....  And makes the
Insight to crash silently without any kind of visible error messages when
trying
to start the non-working windows. Without it these things just don't do
anything,
like crash the program....

 Ok, I had forgot some basic things.... The first is that the DLLs for
Win32s, at
least the 16-bit 'TCL1680.DLL',  must be produced with either Borland C++ 4.5

(or newer for the 32-bit DLLs)  or with MSVC [456] and MSVC 1.5 (for the
16-bit DLL). Perhaps mixing the Mingw-produced 32-bit DLLs with the
Borland/MS-produced 16-bit one could also succeed but I wouldn't be sure
about that....

  The results I have now got, came from using the one year old DLLs where the

'TK80.DLL' was made from the 8.0pl2 sources and with BCC 4.52, just as were
the TCL*80.DLLs, but built from the 8.04 sources....  The 8.04-version
TK80.DLL
didn't seem to work at all when I built the DLLs.  The 98r2-Insight used
8.0pl2 but
the newer 8.04 DLLs seemed to work just as well or better with the 98r2-based

Insights, anyway the newer Insights seem to require the 8.04-versions...
Really a
mess of DLLs now....

 So,  getting working DLLs for Win32s seems to be the biggest problem.  The
tcl/tk/itcl/itk 8.04 and 3.0 sources seem to have support for Borland C++ and

MSVC (besides for GCC), but probably the support needs some fixing, my first
try with BCC 4.52 (again)  produced DLLs which caused all kind of "Unhandled
Exception"- messages...

  I also tried with the quite current 20010501 weekly snapshots and the
resulted
Insight worked quite identically with the 5.0 when using the 5.0 scripts, but
produced
only the frame for the Source Window, no menubar, no toolbar, just the plain
window when using its own scripts....   BTW, I produce  the Insights under
Linux
using Linux-x-Mingw cross-tools for Win32/Win32s  hosts.

 There are things like the need to cut the script and image filenames to 8
chars so
that Win3.1/Win32s can find them, so perhaps there are still some not found /
not
yet renamed files I'm not aware of....  The errors/warnings must first be get
visible.
I fixed the 'verror()' in 'gdb/utils.c' to output the messages via
MessageBox'es, not
via the unexisting 'Win32 console', but this probably wasn't enough because
the
crashes after setting the GDBTK_DEBUG don't cause any messages....  Perhaps
when checking everything,  it starts to work....  One thing I found via the
new
MessageBoxes was that for some reason the function/symbol 'Tcl_Gets' is not
included into the Mingw-produced 8.04 DLLs but the BCC-produced ones have
it....

 The tcl/tk-8.04, itcl-3.0 and tix-4.1 docs say that Win32s is supported and
the DLLs
should work, but I remember finding prebuilt 8.04-DLLs only for Win32
without the
TCL1680.DLL and including the Win32-only MSVCRT.DLL, compiled with MSVC 5.0
....
Anyhow trying the newer BCC 5.2, 5.3, 5.4 and 5.5 compilers for the 32-bit
DLL stuff
instead of the BCC 4.52 could help (or make things worse).  But all this is
just an
experiment,  the Win32s-port may as well remain in that 98r2-level for this
early-90s
opsys....   The tcl/tk-8.1 and newer then aren't supported for Win32s, so
there would
be a natural dead for the port later....

Cheers, Kai

Re: Insight port to the 'original' Windows

[Keith Seitz]

On Thu, 10 May 2001, Kai Ruottu wrote:

>  So,  getting working DLLs for Win32s seems to be the biggest problem.  The
> tcl/tk/itcl/itk 8.04 and 3.0 sources seem to have support for Borland C++ and

Yes, this is the problem. Tcl/Tk stopped support for Win32s when Tk8.0 was
released...

You are probably using "Gdbtk" written by Stu Grossman a couple of years
ago. What we now call "Insight" is a newer version of this same
application, but it requires Tcl/Tk >= 8.0. Sorry.

Keith