fopen doesn't return when opening a fifo!!

[David Bateman]

The following code fragment fails for me on Octave 3.0.x and 3.1.x

tn = tmpnam ();
## 6*8*8 = 0600 mode
[err, msg] = mkfifo(tn, 6*8*8)
gpin = fopen (tn, "r");
fclose (gpin);
unlink (tn);

as the fopen on the fifo does not return. I originally thought this
might be an issue with my libc on my 64-bit system, but the same occurs
on an old 32-bit system. Do others see the same bug? Anyone know what is
going on?

D.

--
David Bateman                                David.Bateman@...
Motorola Labs - Paris                        [phone] (Ph)
Parc Les Algorithmes, Commune de St Aubin    [phone] (Mob)
91193 Gif-Sur-Yvette FRANCE                  [phone] (Fax)

The information contained in this communication has been classified as:

[x] General Business Information
[ ] Motorola Internal Use Only
[ ] Motorola Confidential Proprietary

_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave

[Ben Abbott]

On Mar 31, 2008, at 9:15 AM, David Bateman wrote:

> The following code fragment fails for me on Octave 3.0.x and 3.1.x
>
> tn = tmpnam ();
> ## 6*8*8 = 0600 mode
> [err, msg] = mkfifo(tn, 6*8*8)
> gpin = fopen (tn, "r");
> fclose (gpin);
> unlink (tn);
>
> as the fopen on the fifo does not return. I originally thought this
> might be an issue with my libc on my 64-bit system, but the same  
> occurs
> on an old 32-bit system. Do others see the same bug? Anyone know  
> what is
> going on?
>
> D.

Yes I see the same, and my activity monitor indicates Octave is idle.

Ben
_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave

[Ben Abbott]

On Mar 31, 2008, at 9:49 AM, Ben Abbott wrote:

>
> On Mar 31, 2008, at 9:15 AM, David Bateman wrote:
>> The following code fragment fails for me on Octave 3.0.x and 3.1.x
>>
>> tn = tmpnam ();
>> ## 6*8*8 = 0600 mode
>> [err, msg] = mkfifo(tn, 6*8*8)
>> gpin = fopen (tn, "r");
>> fclose (gpin);
>> unlink (tn);
>>
>> as the fopen on the fifo does not return. I originally thought this
>> might be an issue with my libc on my 64-bit system, but the same
>> occurs
>> on an old 32-bit system. Do others see the same bug? Anyone know
>> what is
>> going on?
>>
>> D.
>
> Yes I see the same, and my activity monitor indicates Octave is idle.
>

 > eval (sprintf ('ls -l %s', tn))
prw------- 1 bpabbott wheel 0 2008-03-31 09:44 /var/tmp/oct-Z5XKZQ

hmmm ... that's not what I expected to see.

 > ls -l /var/tmp/oct-*
prw------- 1 bpabbott wheel   0 2008-03-31 09:43 /var/tmp/oct-9DVwxT
prw------- 1 bpabbott wheel   0 2008-03-31 09:44 /var/tmp/oct-Z5XKZQ
-rw-r--r-- 1 root     wheel   0 2007-11-24 22:25 /var/tmp/oct-p0PLLi
-rw-r--r-- 1 root     wheel   0 2007-12-02 22:47 /var/tmp/oct-pqjfGX
[snip]

*My* temporary oct-files prior to Dec 2 have different permissions (I  
have no idea when the binary that produced them was built). In any  
event, it looks like the permissions have changed.

A quick search of  src/ChangeLog, turns up this.

  2172 2007-11-02  John W. Eaton  <jwe@...>
  2173
  2174         * file-io.cc (fopen_mode_to_ios_mode): Use  
std::ios::app instead
  2175         of std::ios::ate.

I have no idea if this is even relevant, but since I'm stuck at home  
waiting on a tow truck :-( ... I thought I'd poke around.

Ben
_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave

[Jaroslav Hajek-2]

On Mon, Mar 31, 2008 at 3:15 PM, David Bateman
<David.Bateman@...> wrote:

> The following code fragment fails for me on Octave 3.0.x and 3.1.x
>
>  tn = tmpnam ();
>  ## 6*8*8 = 0600 mode
>  [err, msg] = mkfifo(tn, 6*8*8)
>  gpin = fopen (tn, "r");
>  fclose (gpin);
>  unlink (tn);
>
>  as the fopen on the fifo does not return. I originally thought this
>  might be an issue with my libc on my 64-bit system, but the same occurs
>  on an old 32-bit system. Do others see the same bug? Anyone know what is
>  going on?

That's what is supposed to happen, isn't it?

The GNU C library manual, chapter 15.3 "FIFO Special Files" says:
"Opening a FIFO for reading normally blocks until some other process opens
the same FIFO for writing, and vice versa."

I think that you have to open specially if you want to avoid this. In
general, it is a useful behaviour IMHO, as opening a FIFO usually
means listening for commands.

cheers,

--
RNDr. Jaroslav Hajek
computing expert
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz
_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave

[David Bateman]

Jaroslav Hajek wrote:

> On Mon, Mar 31, 2008 at 3:15 PM, David Bateman
> <David.Bateman@...> wrote:
>  
>> The following code fragment fails for me on Octave 3.0.x and 3.1.x
>>
>>  tn = tmpnam ();
>>  ## 6*8*8 = 0600 mode
>>  [err, msg] = mkfifo(tn, 6*8*8)
>>  gpin = fopen (tn, "r");
>>  fclose (gpin);
>>  unlink (tn);
>>
>>  as the fopen on the fifo does not return. I originally thought this
>>  might be an issue with my libc on my 64-bit system, but the same occurs
>>  on an old 32-bit system. Do others see the same bug? Anyone know what is
>>  going on?
>>    
>
>
> That's what is supposed to happen, isn't it?
>
> The GNU C library manual, chapter 15.3 "FIFO Special Files" says:
> "Opening a FIFO for reading normally blocks until some other process opens
> the same FIFO for writing, and vice versa."
>
> I think that you have to open specially if you want to avoid this. In
> general, it is a useful behaviour IMHO, as opening a FIFO usually
> means listening for commands.
>
>  

Ok, then I have some other strange effect. I've tried to convert Petr's

ginput function from

http://gnuplot.sourceforge.net/links.html

to the new graphics handle interface, and I believe I have the code
setup correctly. However, opening the fifo to read the feedback from
gnuplot blocks. The code I currently have is attached if anyone wants to
take a look at this..

D.




--
David Bateman                                David.Bateman@...
Motorola Labs - Paris                        [phone] (Ph)
Parc Les Algorithmes, Commune de St Aubin    [phone] (Mob)
91193 Gif-Sur-Yvette FRANCE                  [phone] (Fax)

The information contained in this communication has been classified as:

[x] General Business Information
[ ] Motorola Internal Use Only
[ ] Motorola Confidential Proprietary


## Copyright (C) 2004, 2006, 2008 Petr Mikulik
##
## This file is part of Octave.
##
## Octave is free software; you can redistribute it and/or modify it
## under the terms of the GNU General Public License as published by
## the Free Software Foundation; either version 3 of the License, or (at
## your option) any later version.
##
## Octave is distributed in the hope that it will be useful, but
## WITHOUT ANY WARRANTY; without even the implied warranty of
## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
## General Public License for more details.
##
## You should have received a copy of the GNU General Public License
## along with Octave; see the file COPYING.  If not, see
## <http://www.gnu.org/licenses/>.

## -*- texinfo -*-
## @deftypefn {Function File} {[@var{x}, @var{y}, @var{buttons}] =} ginput (@var{n})
## Return which mouse buttons were pressed and keys were hit on the current
## figure. If @var{n} is defined, then wait for @var{n} mouse clicks
## before returning. If @var{n} is not defined, then @code{ginput} will
## loop until the return key is pressed.
## @end deftypefn

## This is ginput.m implementation for gnuplot and X11.
## It requires gnuplot 4.1 and later.

## This file initially bore the copyright statement
## Petr Mikulik
## History: June 2006; August 2005; June 2004; April 2004
## License: public domain

function varargout = ginput (n)

  if (nargin > 1)
    print_usage ();
  elseif (nargin == 0)
    ## A startup-value
    n = 100;
  endif

  f = gcf();
  drawnow ();
  backend = (get (f, "__backend__"));

  varargout = cell (1, nargout);
  if (nargin == 0)
    [varargout{:}] = feval (strcat ("__", backend, "_ginput__"), f);
  else
    [varargout{:}] = feval (strcat ("__", backend, "_ginput__"), f, n);
  endif

endfunction

function [x, y, button] = __gnuplot_ginput__ (f, n)

  stream = get (f, "__plot_stream__");

  if (compare_versions (__gnuplot_version__ (), "4.0", "<="))
    error ("ginput: version %s of gnuplot not supported", gnuplot_version ());
  endif

  if (nargin == 0)
    x = zeros (n, 1);
    y = zeros (n, 1);
    button = zeros (n, 1);
  else
    x = zeros (n, 1);
    y = zeros (n, 1);
    button = zeros (n, 1);
  endif

  gpin_name = tmpnam ();

  ## 6*8*8 ==  0600
  [err, msg] = mkfifo(gpin_name, 6*8*8);
  if (err != 0)
    error ("ginput: Can not open fifo (%s)", msg);
  endif    

  k = 0;
  while (true)
    k++;
    fprintf (stream, "set print \"%s\";\n", gpin_name);
    gpin = fopen (gpin_name, "r");
    fprintf (stream, "pause mouse any\n\n");

    ## Notes: MOUSE_* can be undefined if user closes gnuplot by "q" or Alt-F4.
    ## Further, this abrupt close also requires the leading "\n" on the
    ## next line.
    fprintf (stream, "\nif (exists(\"MOUSE_KEY\") && exists(\"MOUSE_X\")) print MOUSE_X, MOUSE_Y, MOUSE_KEY; else print \"0 0 -1\"\n");

    ## close output file, otherwise all blocks (why?!)
    fprintf (stream, "set print\n");

    ## Now read from fifo
    [x(k), y(k), button(k), count] = fscanf (gpin, "%f %f %d", "C");
    fclose (gpin);

    if ([x(k), y(k), button(k)] == [0, 0, -1])
      ## Mousing not active (no plot until now)
      break;
    endif

    if (nargin > 0)
      ## input argument n was given => stop when k==n
      if (k == n)
        break;
      endif
    else
      ## input argument n not given => stop when hitting a return key
      ## if (button(k)==0x0D || button(k)==0x0A)
      ##   ## hit Return or Enter
      if (button(k) == 0x0D)
        ## hit Return
        x (k : end) = [];
        y (k : end) = [];
        button (k : end) = [];
        break;
      endif
    endif
  endwhile

  unlink(gpin_name);
endfunction

_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave

[David Bateman]

David Bateman wrote:

> Jaroslav Hajek wrote:
>  
>> On Mon, Mar 31, 2008 at 3:15 PM, David Bateman
>> <David.Bateman@...> wrote:
>>  
>>    
>>> The following code fragment fails for me on Octave 3.0.x and 3.1.x
>>>
>>>  tn = tmpnam ();
>>>  ## 6*8*8 = 0600 mode
>>>  [err, msg] = mkfifo(tn, 6*8*8)
>>>  gpin = fopen (tn, "r");
>>>  fclose (gpin);
>>>  unlink (tn);
>>>
>>>  as the fopen on the fifo does not return. I originally thought this
>>>  might be an issue with my libc on my 64-bit system, but the same occurs
>>>  on an old 32-bit system. Do others see the same bug? Anyone know what is
>>>  going on?
>>>    
>>>      
>> That's what is supposed to happen, isn't it?
>>
>> The GNU C library manual, chapter 15.3 "FIFO Special Files" says:
>> "Opening a FIFO for reading normally blocks until some other process opens
>> the same FIFO for writing, and vice versa."
>>
>> I think that you have to open specially if you want to avoid this. In
>> general, it is a useful behaviour IMHO, as opening a FIFO usually
>> means listening for commands.
>>
>>  
>>    
> Ok, then I have some other strange effect. I've tried to convert Petr's
> ginput function from
>
> http://gnuplot.sourceforge.net/links.html
>
> to the new graphics handle interface, and I believe I have the code
> setup correctly. However, opening the fifo to read the feedback from
> gnuplot blocks. The code I currently have is attached if anyone wants to
> take a look at this..
>
> D.
>
>
>  

Oppps, ok I figured it out.. I need to flush the stream to gnuplot

otherwise gnuplot doesn't open the fifo and its blocked as you say.. The
attached version of ginput.m works with Octave 3.1.x and versions of
gnuplot newer than 4.1.x.. I'm not sure this works with other terminals
than X11 though so I'm not sure its a generic solution for ginput for
Octave with a gnuplot backend.

Find a attached a working version of ginput.m for Octave 3.1.x. To get
it to work for Octave 3.0, only the test for the __backend__ property
needs to be removed and the __gnuplot_ginput__ subfunction called
unconditionally.

Regards
David

--
David Bateman                                David.Bateman@...
Motorola Labs - Paris                        [phone] (Ph)
Parc Les Algorithmes, Commune de St Aubin    [phone] (Mob)
91193 Gif-Sur-Yvette FRANCE                  [phone] (Fax)

The information contained in this communication has been classified as:

[x] General Business Information
[ ] Motorola Internal Use Only
[ ] Motorola Confidential Proprietary


## Copyright (C) 2004, 2006, 2008 Petr Mikulik
##
## This file is part of Octave.
##
## Octave is free software; you can redistribute it and/or modify it
## under the terms of the GNU General Public License as published by
## the Free Software Foundation; either version 3 of the License, or (at
## your option) any later version.
##
## Octave is distributed in the hope that it will be useful, but
## WITHOUT ANY WARRANTY; without even the implied warranty of
## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
## General Public License for more details.
##
## You should have received a copy of the GNU General Public License
## along with Octave; see the file COPYING.  If not, see
## <http://www.gnu.org/licenses/>.

## -*- texinfo -*-
## @deftypefn {Function File} {[@var{x}, @var{y}, @var{buttons}] =} ginput (@var{n})
## Return which mouse buttons were pressed and keys were hit on the current
## figure. If @var{n} is defined, then wait for @var{n} mouse clicks
## before returning. If @var{n} is not defined, then @code{ginput} will
## loop until the return key is pressed.
## @end deftypefn

## This is ginput.m implementation for gnuplot and X11.
## It requires gnuplot 4.1 and later.

## This file initially bore the copyright statement
## Petr Mikulik
## History: June 2006; August 2005; June 2004; April 2004
## License: public domain

function varargout = ginput (n)

  if (nargin > 1)
    print_usage ();
  elseif (nargin == 0)
    ## A startup-value
    n = 100;
  endif

  f = gcf();
  drawnow ();
  backend = (get (f, "__backend__"));

  varargout = cell (1, nargout);
  if (nargin == 0)
    [varargout{:}] = feval (strcat ("__", backend, "_ginput__"), f);
  else
    [varargout{:}] = feval (strcat ("__", backend, "_ginput__"), f, n);
  endif

endfunction

function [x, y, button] = __gnuplot_ginput__ (f, n)

  stream = get (f, "__plot_stream__");

  if (compare_versions (__gnuplot_version__ (), "4.0", "<="))
    error ("ginput: version %s of gnuplot not supported", gnuplot_version ());
  endif

  if (nargin == 0)
    x = zeros (n, 1);
    y = zeros (n, 1);
    button = zeros (n, 1);
  else
    x = zeros (n, 1);
    y = zeros (n, 1);
    button = zeros (n, 1);
  endif

  gpin_name = tmpnam ();

  ## 6*8*8 ==  0600
  [err, msg] = mkfifo(gpin_name, 6*8*8);
  if (err != 0)
    error ("ginput: Can not open fifo (%s)", msg);
  endif    

  k = 0;
  while (true)
    k++;
    fprintf (stream, "set print \"%s\";\n", gpin_name);
    fflush (stream);
    gpin = fopen (gpin_name, "r");
    fprintf (stream, "pause mouse any\n\n");

    ## Notes: MOUSE_* can be undefined if user closes gnuplot by "q" or Alt-F4.
    ## Further, this abrupt close also requires the leading "\n" on the
    ## next line.
    fprintf (stream, "\nif (exists(\"MOUSE_KEY\") && exists(\"MOUSE_X\")) print MOUSE_X, MOUSE_Y, MOUSE_KEY; else print \"0 0 -1\"\n");

    ## close output file, otherwise all blocks (why?!)
    fprintf (stream, "set print\n");
    fflush (stream);

    ## Now read from fifo
    [x(k), y(k), button(k), count] = fscanf (gpin, "%f %f %d", "C");
    fclose (gpin);

    if ([x(k), y(k), button(k)] == [0, 0, -1])
      ## Mousing not active (no plot until now)
      break;
    endif

    if (nargin > 0)
      ## input argument n was given => stop when k==n
      if (k == n)
        break;
      endif
    else
      ## input argument n not given => stop when hitting a return key
      ## if (button(k)==0x0D || button(k)==0x0A)
      ##   ## hit Return or Enter
      if (button(k) == 0x0D)
        ## hit Return
        x (k : end) = [];
        y (k : end) = [];
        button (k : end) = [];
        break;
      endif
    endif
  endwhile

  unlink(gpin_name);
endfunction

_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave

[John W. Eaton]

On 31-Mar-2008, David Bateman wrote:

| Oppps, ok I figured it out.. I need to flush the stream to gnuplot
| otherwise gnuplot doesn't open the fifo and its blocked as you say.. The
| attached version of ginput.m works with Octave 3.1.x and versions of
| gnuplot newer than 4.1.x.. I'm not sure this works with other terminals
| than X11 though so I'm not sure its a generic solution for ginput for
| Octave with a gnuplot backend.
|
| Find a attached a working version of ginput.m for Octave 3.1.x. To get
| it to work for Octave 3.0, only the test for the __backend__ property
| needs to be removed and the __gnuplot_ginput__ subfunction called
| unconditionally.

Since it seems to work, I checked it in.  I split the
__gnuplot_ginput__ function into a separate file and added an
unwind_protect block to ensure that the tmp file is cleaned up.  I'm
sure we'll find out soon enough whether it works with non-X11
terminals...

Thanks,

jwe
_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave

[dbateman3]

John W. Eaton wrote:

> On 31-Mar-2008, David Bateman wrote:
>
> | Oppps, ok I figured it out.. I need to flush the stream to gnuplot
> | otherwise gnuplot doesn't open the fifo and its blocked as you say.. The
> | attached version of ginput.m works with Octave 3.1.x and versions of
> | gnuplot newer than 4.1.x.. I'm not sure this works with other terminals
> | than X11 though so I'm not sure its a generic solution for ginput for
> | Octave with a gnuplot backend.
> |
> | Find a attached a working version of ginput.m for Octave 3.1.x. To get
> | it to work for Octave 3.0, only the test for the __backend__ property
> | needs to be removed and the __gnuplot_ginput__ subfunction called
> | unconditionally.
>
> Since it seems to work, I checked it in.  I split the
> __gnuplot_ginput__ function into a separate file and added an
> unwind_protect block to ensure that the tmp file is cleaned up.  I'm
> sure we'll find out soon enough whether it works with non-X11
> terminals...
>

It doesn't work for Windows due to the use of mkfifo. Though probably
works on Macs. I have some mods, and the waitforbuttonpress and gtext
function as well to come.. I'm trying to figure out a clean way to get
this code to work on systems with mkfifo as well with char files, though
 I'm having problems with blocking reads from the chat file or to the
gnuplot stream to allow the chat file to be synchronized between gnuplot
and Octave. I'll try and send a patch soon.

D.
_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave

[John W. Eaton]

On 31-Mar-2008, David Bateman wrote:

| It doesn't work for Windows due to the use of mkfifo. Though probably
| works on Macs. I have some mods, and the waitforbuttonpress and gtext
| function as well to come.. I'm trying to figure out a clean way to get
| this code to work on systems with mkfifo as well with char files, though
|  I'm having problems with blocking reads from the chat file or to the
| gnuplot stream to allow the chat file to be synchronized between gnuplot
| and Octave. I'll try and send a patch soon.

If we can make these functions work then OK, but I wouldn't work too
hard at it.  I expect they will be obsolete once the OpenGL code is
the default.

jwe
_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave

[dbateman3]

John W. Eaton wrote:

> On 31-Mar-2008, David Bateman wrote:
>
> | It doesn't work for Windows due to the use of mkfifo. Though probably
> | works on Macs. I have some mods, and the waitforbuttonpress and gtext
> | function as well to come.. I'm trying to figure out a clean way to get
> | this code to work on systems with mkfifo as well with char files, though
> |  I'm having problems with blocking reads from the chat file or to the
> | gnuplot stream to allow the chat file to be synchronized between gnuplot
> | and Octave. I'll try and send a patch soon.
>
> If we can make these functions work then OK, but I wouldn't work too
> hard at it.  I expect they will be obsolete once the OpenGL code is
> the default.
>
> jwe

Well I have the code to get around the mkfifo issue in Windows using a
chat file. However the gnuplot command "pause mouse any" doesn't work in
the Windows terminal..

In any case here is a patch that adds the waitforbuttonpress and gtext
commands, fixes a few bugs in __gnuplot_ginput__ and adds a work around
for missing mkfifo functions. Note I just assume that if ispc() is true
then we don't have the mkfifo function. I suppose we should do something
cleaner than that. We'll soon find out if these work on Macs I imagine :-)

D.


# HG changeset patch
# User David Bateman <dbateman@...>
# Date 1207000141 -7200
# Node ID fb76ce55d6bd4c249cab45b7ee8a82c1027f519c
# Parent  3523775b196695ac07913009b940ea1f573dc35d
Add gtext, waitforbuttonpress. Attempt to get ginput working under windows

diff --git a/scripts/ChangeLog b/scripts/ChangeLog
--- a/scripts/ChangeLog
+++ b/scripts/ChangeLog
@@ -1,5 +1,12 @@ 2008-03-31  David Bateman  <dbateman@fre
 2008-03-31  David Bateman  <dbateman@...>
 
+ * plot/gtext.m: New function to place text on a plot.
+ * plot/waitforbuttonpress.m: New function.
+ * plot/Makefile.in (SOURCES): Add them to the list.
+ * plot/__gnuplot_ginput__.m: Bug fix for nargin==1. Workaround for
+ missing mkfifo under windows.
+ * plot/ginput.m: Eliminate setting og n.
+
  * plot/newplot.m: Treat the case of plotyy.
  * plot/__go_draw_axes__.m: Set the tick direction in the main call
  the set tics rather than separately to avoid issues with multiple
diff --git a/scripts/plot/Makefile.in b/scripts/plot/Makefile.in
--- a/scripts/plot/Makefile.in
+++ b/scripts/plot/Makefile.in
@@ -110,6 +110,7 @@ SOURCES = \
   gcf.m \
   ginput.m \
   grid.m \
+  gtext.m \
   hidden.m \
   hist.m \
   hold.m \
@@ -165,6 +166,7 @@ SOURCES = \
   text.m \
   title.m \
   view.m \
+  waitforbuttonpress.m \
   xlabel.m \
   xlim.m \
   ylabel.m \
diff --git a/scripts/plot/__gnuplot_ginput__.m b/scripts/plot/__gnuplot_ginput__.m
--- a/scripts/plot/__gnuplot_ginput__.m
+++ b/scripts/plot/__gnuplot_ginput__.m
@@ -16,6 +16,11 @@
 ## along with Octave; see the file COPYING.  If not, see
 ## <http://www.gnu.org/licenses/>.
 
+## -*- texinfo -*-
+## @deftypefn {Function File} {[@var{x}, @var{y}, @var{buttons}] =} __gnuplot_ginput__ (@var{f}, @var{n})
+## Undocumented internal function.
+## @end deftypefn
+
 ## This is ginput.m implementation for gnuplot and X11.
 ## It requires gnuplot 4.1 and later.
 
@@ -24,12 +29,9 @@
 ## History: June 2006; August 2005; June 2004; April 2004
 ## License: public domain
 
-## -*- texinfo -*-
-## @deftypefn {Function File} {[@var{x}, @var{y}, @var{buttons}] =} __gnuplot_ginput__ (@var{f}, @var{n})
-## Undocumented internal function.
-## @end deftypefn
+function [x, y, button] = __gnuplot_ginput__ (f, n)
 
-function [x, y, button] = __gnuplot_ginput__ (f, n)
+  persistent have_mkfifo = ! ispc ();
 
   stream = get (f, "__plot_stream__");
 
@@ -37,10 +39,10 @@ function [x, y, button] = __gnuplot_ginp
     error ("ginput: version %s of gnuplot not supported", gnuplot_version ());
   endif
 
-  if (nargin == 0)
-    x = zeros (n, 1);
-    y = zeros (n, 1);
-    button = zeros (n, 1);
+  if (nargin == 1)
+    x = zeros (100, 1);
+    y = zeros (100, 1);
+    button = zeros (100, 1);
   else
     x = zeros (n, 1);
     y = zeros (n, 1);
@@ -49,11 +51,14 @@ function [x, y, button] = __gnuplot_ginp
 
   gpin_name = tmpnam ();
 
-  ## 6*8*8 ==  0600
-  [err, msg] = mkfifo (gpin_name, 6*8*8);
-  if (err != 0)
-    error ("ginput: Can not open fifo (%s)", msg);
-  endif    
+  if (have_mkfifo)
+    ## Use pipes if not on Windows. Mode: 6*8*8 ==  0600
+    [err, msg] = mkfifo(gpin_name, 6*8*8);
+
+    if (err != 0)
+      error ("ginput: Can not open fifo (%s)", msg);
+    endif
+  endif
 
   unwind_protect
 
@@ -62,7 +67,12 @@ function [x, y, button] = __gnuplot_ginp
       k++;
       fprintf (stream, "set print \"%s\";\n", gpin_name);
       fflush (stream);
-      gpin = fopen (gpin_name, "r");
+      if (have_mkfifo)
+ [gpin, err] = fopen (gpin_name, "r");
+ if (err != 0)
+  error ("ginput: Can not open fifo (%s)", msg);
+ endif
+      endif
       fputs (stream, "pause mouse any;\n\n");
 
       ## Notes: MOUSE_* can be undefined if user closes gnuplot by "q"
@@ -74,8 +84,27 @@ function [x, y, button] = __gnuplot_ginp
       fputs (stream, "set print;\n");
       fflush (stream);
 
-      ## Now read from fifo.
-      [x(k), y(k), button(k), count] = fscanf (gpin, "%f %f %d", "C");
+      if (! have_mkfifo)
+ while (exist (gpin_name, "file") == 0)
+ endwhile
+ [gpin, msg] = fopen (gpin_name, "r");
+
+ if (gpin < 0)
+  error ("ginput: Can not open file (%s)", msg);
+ endif
+
+ ## Now read from file
+ count = 0;
+ while (count == 0)
+  [xk, yk, buttonk, count] = fscanf (gpin, "%f %f %d", "C");
+ endwhile
+ x(k) = xk;
+ y(k) = yk;
+ button (k) = buttonk;
+      else
+ ## Now read from fifo.
+ [x(k), y(k), button(k), count] = fscanf (gpin, "%f %f %d", "C");
+      endif
       fclose (gpin);
 
       if ([x(k), y(k), button(k)] == [0, 0, -1])
@@ -83,7 +112,7 @@ function [x, y, button] = __gnuplot_ginp
  break;
       endif
 
-      if (nargin > 0)
+      if (nargin > 1)
  ## Input argument n was given => stop when k == n.
  if (k == n)
   break;
diff --git a/scripts/plot/ginput.m b/scripts/plot/ginput.m
--- a/scripts/plot/ginput.m
+++ b/scripts/plot/ginput.m
@@ -28,9 +28,6 @@ function varargout = ginput (n)
 
   if (nargin > 1)
     print_usage ();
-  elseif (nargin == 0)
-    ## A startup-value
-    n = 100;
   endif
 
   f = gcf ();
diff --git a/scripts/plot/gtext.m b/scripts/plot/gtext.m
new file mode 100644
--- /dev/null
+++ b/scripts/plot/gtext.m
@@ -0,0 +1,53 @@
+## Copyright (C) 2008  David Bateman
+##
+## This file is part of Octave.
+##
+## Octave is free software; you can redistribute it and/or modify it
+## under the terms of the GNU General Public License as published by
+## the Free Software Foundation; either version 3 of the License, or (at
+## your option) any later version.
+##
+## Octave is distributed in the hope that it will be useful, but
+## WITHOUT ANY WARRANTY; without even the implied warranty of
+## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+## General Public License for more details.
+##
+## You should have received a copy of the GNU General Public License
+## along with Octave; see the file COPYING.  If not, see
+## <http://www.gnu.org/licenses/>.
+
+## -*- texinfo -*-
+## @deftypefn {Function File} {} gtext (@var{s})
+## @deftypefnx {Function File} {} gtext (@dots{}, @var{prop}, @var{val})
+## Places text on the current figure. The text can be defined by the
+## string @var{s}. If @var{s} is a cell array, each element of the cell
+## array is written to a separate line.
+##
+## Additional arguments are passed to the underlying text object as
+## properties.
+## @seealso{ginput}
+## @end deftypefn
+
+function gtext (s, varargin)
+
+  if (nargin < 1)
+    print_usage ();
+  endif
+
+  if (iscell (s))
+    s = s(:);
+    if (all (cellfun (@ischar, s)))
+      tmp(1:2:2*length(s)) = s;
+      tmp(2:2:2*length(s)) = "\n";
+      s = strcat (tmp{:});
+    else
+      error ("gtext: expecting a string or cell array of strings");
+    endif
+  elseif (!ischar (s))
+    error ("gtext: expecting a string or cell array of strings");
+  endif
+
+  [x, y] = ginput (1);
+  text (x, y, s, varargin{:});
+endfunction
+
diff --git a/scripts/plot/waitforbuttonpress.m b/scripts/plot/waitforbuttonpress.m
new file mode 100644
--- /dev/null
+++ b/scripts/plot/waitforbuttonpress.m
@@ -0,0 +1,46 @@
+## Copyright (C) 2004, 2006, 2008 Petr Mikulik
+##
+## This file is part of Octave.
+##
+## Octave is free software; you can redistribute it and/or modify it
+## under the terms of the GNU General Public License as published by
+## the Free Software Foundation; either version 3 of the License, or (at
+## your option) any later version.
+##
+## Octave is distributed in the hope that it will be useful, but
+## WITHOUT ANY WARRANTY; without even the implied warranty of
+## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+## General Public License for more details.
+##
+## You should have received a copy of the GNU General Public License
+## along with Octave; see the file COPYING.  If not, see
+## <http://www.gnu.org/licenses/>.
+
+## -*- texinfo -*-
+## @deftypefn {Function File} {@var{b} =} waitforbuttonpress ()
+## Wait for button or mouse press.over a figure window. The value of
+## @var{b} returns 0 if a mouse button was pressed or 1 is a key was
+## pressed.
+## @seealso{ginput}
+## @end deftypefn
+
+## The original version of this code bore the copyright
+## Author: Petr Mikulik
+## License: public domain
+
+function a = waitforbuttonpress ()
+
+  if (nargin != 0 || nargout > 1)
+    print_usage ();
+  endif
+
+  [x, y, k] = ginput(1);
+
+  if (nargout == 1)
+    if (k <= 5)
+      a = 0;
+    else
+      a = 1;
+    endif
+  endif
+endfunction

_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave

[John W. Eaton]

On 31-Mar-2008, David Bateman wrote:

| In any case here is a patch that adds the waitforbuttonpress and gtext
| commands, fixes a few bugs in __gnuplot_ginput__ and adds a work around
| for missing mkfifo functions. Note I just assume that if ispc() is true
| then we don't have the mkfifo function. I suppose we should do something
| cleaner than that. We'll soon find out if these work on Macs I imagine :-)

OK.  Before I add it and commit, is there a reason to prefer

| +  if (iscell (s))
| +    s = s(:);
| +    if (all (cellfun (@ischar, s)))
| +      tmp(1:2:2*length(s)) = s;
| +      tmp(2:2:2*length(s)) = "\n";
| +      s = strcat (tmp{:});

over

  if (iscellstr (s))
    s = sprintf ("%s\n", s{:});
  endif

?

jwe

_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave

[dbateman3]

John W. Eaton wrote:

> On 31-Mar-2008, David Bateman wrote:
>
> | In any case here is a patch that adds the waitforbuttonpress and gtext
> | commands, fixes a few bugs in __gnuplot_ginput__ and adds a work around
> | for missing mkfifo functions. Note I just assume that if ispc() is true
> | then we don't have the mkfifo function. I suppose we should do something
> | cleaner than that. We'll soon find out if these work on Macs I imagine :-)
>
> OK.  Before I add it and commit, is there a reason to prefer
>
> | +  if (iscell (s))
> | +    s = s(:);
> | +    if (all (cellfun (@ischar, s)))
> | +      tmp(1:2:2*length(s)) = s;
> | +      tmp(2:2:2*length(s)) = "\n";
> | +      s = strcat (tmp{:});
>
> over
>
>   if (iscellstr (s))
>     s = sprintf ("%s\n", s{:});
>   endif
>
> ?
>
> jwe

No reason. Your version is better..

D.

_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave

[Petr Mikulik]

> > | Oppps, ok I figured it out.. I need to flush the stream to gnuplot
> > | otherwise gnuplot doesn't open the fifo and its blocked as you say.. The
> > | attached version of ginput.m works with Octave 3.1.x and versions of
> > | gnuplot newer than 4.1.x.

Great! I've downloaded ginput.m from http://www.octave.org/hg/octave and
have it working in Octave 3.0. It works fine.

Some comments:

- You can remove this:
## This file initially bore the copyright statement
## Petr Mikulik
## History: June 2006; August 2005; June 2004; April 2004
## License: public domain

- Remove this code, otherwise  ginput  (without input arguments)
  does not work:

  else
    x = zeros (n, 1);
    y = zeros (n, 1);
    button = zeros (n, 1);

- > the gnuplot command "pause mouse any" doesn't work in
  > the Windows terminal..
  It works correctly.

- The Windows binary distribution by Tatsuro MATSUOKA prefers to be run
  under cygwin, and the recommended terminal is X11.
  Does that mean that within cygwin's bash, mkfifo() in Octave may work?
  Or is mkfifo() not at all available in Octave's binary for MSW?

---
Petr Mikulik
_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave

[John W. Eaton]

On  1-Apr-2008, David Bateman wrote:

| John W. Eaton wrote:
| > On 31-Mar-2008, David Bateman wrote:
| >
| > | In any case here is a patch that adds the waitforbuttonpress and gtext
| > | commands, fixes a few bugs in __gnuplot_ginput__ and adds a work around
| > | for missing mkfifo functions. Note I just assume that if ispc() is true
| > | then we don't have the mkfifo function. I suppose we should do something
| > | cleaner than that. We'll soon find out if these work on Macs I imagine :-)
| >
| > OK.  Before I add it and commit, is there a reason to prefer
| >
| > | +  if (iscell (s))
| > | +    s = s(:);
| > | +    if (all (cellfun (@ischar, s)))
| > | +      tmp(1:2:2*length(s)) = s;
| > | +      tmp(2:2:2*length(s)) = "\n";
| > | +      s = strcat (tmp{:});
| >
| > over
| >
| >   if (iscellstr (s))
| >     s = sprintf ("%s\n", s{:});
| >   endif
| >
| > ?
| >
| > jwe
|
| No reason. Your version is better..

OK.  I applied the patch with this change and a few more to handle
empty cell and string arguments.

Thanks,

jwe
_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave

[David Bateman]

Petr Mikulik wrote:

>>> | Oppps, ok I figured it out.. I need to flush the stream to gnuplot
>>> | otherwise gnuplot doesn't open the fifo and its blocked as you say.. The
>>> | attached version of ginput.m works with Octave 3.1.x and versions of
>>> | gnuplot newer than 4.1.x.
>>>      
>
> Great! I've downloaded ginput.m from http://www.octave.org/hg/octave and
> have it working in Octave 3.0. It works fine.
>
> Some comments:
>
> - You can remove this:
> ## This file initially bore the copyright statement
> ## Petr Mikulik
> ## History: June 2006; August 2005; June 2004; April 2004
> ## License: public domain
>  

I don't touch license statements when porting code. But sure we can
remove it if you're ok with that.

> - Remove this code, otherwise  ginput  (without input arguments)
>   does not work:
>
>   else
>     x = zeros (n, 1);
>     y = zeros (n, 1);
>     button = zeros (n, 1);
>  
Yes, I noticed and fixed that.. The fact is the version of ginput.m hat
John originally committed was sent in the couple of minutes before I
packed up and left, and as it was the first working version I wanted to
get it out there.. The latest version in the repository should be fine.

> - > the gnuplot command "pause mouse any" doesn't work in
>   > the Windows terminal..
>   It works correctly.
>  
I tried these functions with Michael's MSVC binary using gnuplot 4.2.2
and had issues. I had to modify ginput.m to force the "backend" variable
to be "gnuplot" as this property doesn't exist in Octave 3.0. However,
it caused some issues, and the ginput command locked up. I then tried to
use "pause mouse any" directly from gnuplot without using Octave at all
with the same result.. Were there some mouse support fixes for the
Windows terminal in gnuplot 4.2.3? Did you test this with Windows?

Basically, I think these function must work under the default terminal
types on Windows, Unix and Mac systems to be truly useful. And so hope
that with an upgrade to gnuplot 4.2.3 that this is possible.

> - The Windows binary distribution by Tatsuro MATSUOKA prefers to be run
>   under cygwin, and the recommended terminal is X11.
>   Does that mean that within cygwin's bash, mkfifo() in Octave may work?
>   Or is mkfifo() not at all available in Octave's binary for MSW?
>  
Doing a quick google search for "cygwin mkfifo" and found

http://lists.mplayerhq.hu/pipermail/mplayer-users/2007-July/068030.html

so cygwin doesn't implement the mkfifo function, or rather it does but
as a dummy function.  So I don't expect the mkfifo will work with
cygwin. That's why I was happy enough with the test "have_mkfifo =
!ispc();"

Regards
David

> ---
> Petr Mikulik
>
>  


--
David Bateman                                David.Bateman@...
Motorola Labs - Paris                        [phone] (Ph)
Parc Les Algorithmes, Commune de St Aubin    [phone] (Mob)
91193 Gif-Sur-Yvette FRANCE                  [phone] (Fax)

The information contained in this communication has been classified as:

[x] General Business Information
[ ] Motorola Internal Use Only
[ ] Motorola Confidential Proprietary

_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave

[Petr Mikulik]

> > - > the gnuplot command "pause mouse any" doesn't work in
> >   > the Windows terminal..
> >   It works correctly.
> >  
> I tried these functions with Michael's MSVC binary using gnuplot 4.2.2
> and had issues. I had to modify ginput.m to force the "backend" variable
> to be "gnuplot" as this property doesn't exist in Octave 3.0. However,
> it caused some issues, and the ginput command locked up. I then tried to
> use "pause mouse any" directly from gnuplot without using Octave at all
> with the same result.. Were there some mouse support fixes for the
> Windows terminal in gnuplot 4.2.3? Did you test this with Windows?
>
> Basically, I think these function must work under the default terminal
> types on Windows, Unix and Mac systems to be truly useful. And so hope
> that with an upgrade to gnuplot 4.2.3 that this is possible.

They do work correctly. I've tried the Windows binary from the gnuplot's
Download section, then:
        plot x
        pause mouse any ... now click or hit a key
        show var

> I tried these functions with Michael's MSVC binary using gnuplot 4.2.2

I don't know this binary. Wherefrom have you got it?

---
Petr Mikulik
_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave

[David Bateman]

Petr Mikulik wrote:
> They do work correctly. I've tried the Windows binary from the gnuplot's
> Download section, then:
> plot x
> pause mouse any ... now click or hit a key
> show var
>
>  
I presume that is a 4.2.3 binary. In any case that is good news.

>> I tried these functions with Michael's MSVC binary using gnuplot 4.2.2
>>    
>
> I don't know this binary. Wherefrom have you got it?
>
>  
This is the windows binary installer of Octave from sourceforge. This is
what the vast majority of Windows users are installing and it is
pre-packaged with a version of gnuplot. It can be found at

http://sourceforge.net/project/showfiles.php?group_id=2888&package_id=40078

Thinking about the issue of how to get around the lack of  mkfifo on
Windows, why can't we use the existing stream that is open to the
gnuplot process to pass the data back from gnuplot to Octave rather than
opening another fifo or chat file? This would mean that the plot_stream
would be opened with popen2 rather than popen and we'd have both an
input and output stream. The major advantage of this is that there is no
loop probing the contents of a chat file under windows that causes
Octave to run at 100% during a ginput operation.

However, one issue I see is that any output that going to the output
stream needs to be read by Octave, where currently it just ends up on
the stdout.. So we'd have to tag the data we want to have gnuplot print
and get accepted by Octave with a tag of some sort, so that we can find
the date we want and throw away the rest.

Also how do we store the input and output plot stream descriptors.. At
the moment there is a single property __plot_stream__ in the figure and
its used for the pipe opened with popen. Do we make that a two element
matrix, or add a new property? What would other backends expect?

I'm tempted to try this modification to Octave out as the advantages
seem to be large for Windows users.

D.

--
David Bateman                                David.Bateman@...
Motorola Labs - Paris                        [phone] (Ph)
Parc Les Algorithmes, Commune de St Aubin    [phone] (Mob)
91193 Gif-Sur-Yvette FRANCE                  [phone] (Fax)

The information contained in this communication has been classified as:

[x] General Business Information
[ ] Motorola Internal Use Only
[ ] Motorola Confidential Proprietary

_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave

[David Bateman]

David Bateman wrote:

> Petr Mikulik wrote:
>  
>> They do work correctly. I've tried the Windows binary from the gnuplot's
>> Download section, then:
>> plot x
>> pause mouse any ... now click or hit a key
>> show var
>>
>>  
>>    
> I presume that is a 4.2.3 binary. In any case that is good news.
>  

I forgot to mention the MSVC binary uses the WXT terminal.. Does "pause
mouse any" work with this terminal? Or did you test only with the X11
terminal under Windows?

D.


--
David Bateman                                David.Bateman@...
Motorola Labs - Paris                        [phone] (Ph)
Parc Les Algorithmes, Commune de St Aubin    [phone] (Mob)
91193 Gif-Sur-Yvette FRANCE                  [phone] (Fax)

The information contained in this communication has been classified as:

[x] General Business Information
[ ] Motorola Internal Use Only
[ ] Motorola Confidential Proprietary

_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave

[Petr Mikulik]

>>> They do work correctly. I've tried the Windows binary from the gnuplot's
>>> Download section, then:
>>>      plot x
>>>      pause mouse any         ... now click or hit a key
>>>      show var
>>>    
>> I presume that is a 4.2.3 binary.

yes, the Windows terminal supports "pause mouse" correctly

> I forgot to mention the MSVC binary uses the WXT terminal.. Does "pause
> mouse any" work with this terminal? Or did you test only with the X11
> terminal under Windows?

OK, now it's more clear. I've tried this MSVC gnuplot under Wine on Linux,
and it seems "pause mouse" does not work on WXT terminal on Windows (on
Linux, it's OK).

So it's a question to Michael Goffioul to fix this issue with the author of
WXT.

BTW, I would expect some contact information in the binary packager
octave-3.0.0-setup.exe (e.g. in README.txt), however, there are no
information whom to contact. I've found the Michael's contact according to
his email to gnuplot mailing list.

---
Petr Mikulik
_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave

[Tatsuro MATSUOKA-2]

Hello Petr and an original questioner

Current pnguplot by oridinariy distubuted does not work at all for octave for windows because octave
send binary data via pipe to set terminal color.

Hi Petr
Do not you really know Michael's pgnuplot ?

It is bundled to the distirbution octave-MSVC which can be dowmload from
http://sourceforge.net/project/showfiles.php?group_id=2888

Please see my site
http://www.geocities.jp/tmoctwin/octmingw.html
and see ReadmeMingwOct3.0.xx.txt

In that page,
please see,

==================================================
2.1.1 gnuplot bundled with MSVC-octave
==================================================


This pgnuplot is known amang the gnuplot deveplopers.
I have once done the discussion about Michael's gnuplot with HBB.

This is a cosole mode pgnuplot for octave for windows.
Its version is 4.2.2.
I think that Michael will upgrade gnuplot for next release.

If the questioner do not wait until next release, please use gnuplot 4.3 (cygwin) introduced in my
page.

http://www.geocities.jp/tmoctwin/

Regards

Tatsuro


--- Petr Mikulik <mikulik@...> wrote:

> > > - > the gnuplot command "pause mouse any" doesn't work in
> > >   > the Windows terminal..
> > >   It works correctly.
> > >  
> > I tried these functions with Michael's MSVC binary using gnuplot 4.2.2
> > and had issues. I had to modify ginput.m to force the "backend" variable
> > to be "gnuplot" as this property doesn't exist in Octave 3.0. However,
> > it caused some issues, and the ginput command locked up. I then tried to
> > use "pause mouse any" directly from gnuplot without using Octave at all
> > with the same result.. Were there some mouse support fixes for the
> > Windows terminal in gnuplot 4.2.3? Did you test this with Windows?
> >
> > Basically, I think these function must work under the default terminal
> > types on Windows, Unix and Mac systems to be truly useful. And so hope
> > that with an upgrade to gnuplot 4.2.3 that this is possible.
>
> They do work correctly. I've tried the Windows binary from the gnuplot's
> Download section, then:
> plot x
> pause mouse any ... now click or hit a key
> show var
>
> > I tried these functions with Michael's MSVC binary using gnuplot 4.2.2
>
> I don't know this binary. Wherefrom have you got it?
>
> ---
> Petr Mikulik
> _______________________________________________
> Bug-octave mailing list
> Bug-octave@...
> https://www.cae.wisc.edu/mailman/listinfo/bug-octave
>

--------------------------------------
Easy + Joy + Powerful = Yahoo! Bookmarks x Toolbar
http://pr.mail.yahoo.co.jp/toolbar/
_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www.cae.wisc.edu/mailman/listinfo/bug-octave