octave-dev

[OctDev] Octave/Win32 new binary package (2.9.12)

[Michael G. <mic...@gm...>]

Hi,

A new binary package of Octave for Win32, based on 2.9.12 is available on
http://www.dbateman.org/?page=octave&lang=EN. Up to now, it's more a
release candidate than an official package. I'd like to find a few people
willing to give it a try and report problems, such that it can maybe become
more official after fixing (it's based on official releases of octave
and octave-forge).
Anyone interested? Thanks.

Michael.

Re: [OctDev] Octave/Win32 new binary package (2.9.12)

[Lukasz K. <lu...@cz...>]

I have tested 2.9.9+ and now 2.9.12 under WinXP Home and Professional
and have following remarks:

- closing window (not typing quit, but simply closing) always crashes
octave and tries to send a report to Microsoft.

- the new panelized terminal has on my both computers a typewriter speed
which is annoying. I must use the raw version on black window.

- there would be nice idea to use rxvt as in 2.1.73. I have tried to
make a shortcut (I have 2.1.73 without cygwin in C:\octave and new
version in C:\octave29) but with no success.

Besides that, this version works quick and no other problems were noticed.

Regards,

Luke

Re: [OctDev] Octave/Win32 new binary package (2.9.12)

[dbateman <dba...@fr...>]

Lukasz Komsta-2 wrote:
> 
> Besides that, this version works quick and no other problems were noticed.
> 

What about the graphics? Michael made his java graphics the default in this
version and I imagine some feedback there would be most appreciated :-)

D.
-- 
View this message in context: http://www.nabble.com/Octave-Win32-new-binary-package-%282.9.12%29-tf3830518.html#a10862476
Sent from the octave-dev mailing list archive at Nabble.com.

Re: [OctDev] Octave/Win32 new binary package (2.9.12)

[Michael G. <mic...@gm...>]

On 5/29/07, Lukasz Komsta <lu...@cz...> wrote:
>
> I have tested 2.9.9+ and now 2.9.12 under WinXP Home and Professional
> and have following remarks:
>
> - closing window (not typing quit, but simply closing) always crashes
> octave and tries to send a report to Microsoft.

I know. I still have to address it.

> - the new panelized terminal has on my both computers a typewriter speed
> which is annoying. I must use the raw version on black window.

This is a Console2 issue. I personally don't use Console2, because of the
slowness feeling. Plain windows command prompt has better responsiveness
and you can still tune its color and size (and use the mouse for copy
and paste).

> - there would be nice idea to use rxvt as in 2.1.73. I have tried to
> make a shortcut (I have 2.1.73 without cygwin in C:\octave and new
> version in C:\octave29) but with no success.

MSVC compiled readline library has problem running within rxvt. When
I start octave within rxvt, the prompt does not show up and the arrow keys
do not work.

Thanks for your report.

Michael.

Re: [OctDev] Octave/Win32 new binary package (2.9.12)

[dbateman <dba...@fr...>]

Michael Goffioul-2 wrote:
> 
> On 5/29/07, Lukasz Komsta <lu...@cz...> wrote:
>>
>> I have tested 2.9.9+ and now 2.9.12 under WinXP Home and Professional
>> and have following remarks:
>>
>> - closing window (not typing quit, but simply closing) always crashes
>> octave and tries to send a report to Microsoft.
> 
> I know. I still have to address it.
> 

For what it is worth I always had the same issue under mingw and never
figured it out... There might even be some messages about it with additional
information, particularly from Paul Kienzle, on the maintainers list in
2005..



>> - the new panelized terminal has on my both computers a typewriter speed
>> which is annoying. I must use the raw version on black window.
> 
> This is a Console2 issue. I personally don't use Console2, because of the
> slowness feeling. Plain windows command prompt has better responsiveness
> and you can still tune its color and size (and use the mouse for copy
> and paste).
> 

Under vmware its even worse .... 



>> - there would be nice idea to use rxvt as in 2.1.73. I have tried to
>> make a shortcut (I have 2.1.73 without cygwin in C:\octave and new
>> version in C:\octave29) but with no success.
> 
> MSVC compiled readline library has problem running within rxvt. When
> I start octave within rxvt, the prompt does not show up and the arrow keys
> do not work.
> 
> 

The rxvt idea is not a good one however. The problem is using the cygwin
rxvt means instaling cygwin, and whats the point when we just got rid of it
with the MSVC build. However the pty emulation in the msys rxvt is seriously
broken and I don't believe it has been fixed since I experimented with this
in 2005. Due to this readline won't work in the msys rxvt and Octave will be
seriously crippled. The windows cmd.exe is therefore really not too bad. 

D.

-- 
View this message in context: http://www.nabble.com/Octave-Win32-new-binary-package-%282.9.12%29-tf3830518.html#a10862655
Sent from the octave-dev mailing list archive at Nabble.com.

Re: [OctDev] Octave/Win32 new binary package (2.9.12)

[Michael G. <mic...@gm...>]

On 5/30/07, Olli Saarela <Oll...@kc...> wrote:
> 1a) Zooming with mouse (rubberband with left button & click with right)
> seems to always operate on the last subplot, no matter which subplot
> you click. This can be tested with
> subplot(311);plot(rand(5,1));subplot(312);plot(rand(5,1));subplot(313);pl=
ot(rand(5,1))

Indeed. Mouse clicking only updates current figure; current axes is not
updated yet. The java graphics backend is still in development and there
are lot of things that do not work correctly. In the final binary package
iteration, I might decide to not put it as default backend (seems wiser).

> 1b) The edit command tries to create a new file to
> fullfile(default_home, "octave"), even though
> the directory doesn't exist
> octave.exe:32> edit new_file.m
> error: edit: could not create C:\Ojs\octave\new_file.m

This also occurs under UNIX. When creating a new file, "edit" script will
put default license statement at the beginning; hence it needs a location
to write the initial file. I guess this is something to fix in the miscella=
neous
package, for instance by creating the target directory if it does not exist=
.

> 1c) Handling of national characters. If I have a directory named l=E4mp=
=F6tila
> (containing an a umlaut and an o umlaut), the command prompt in Console
> shows it in the Windows fashion (correct national characters):
> C:\tmp\test>dir /b
> l=E4mp=F6tila
> but Octave in a Console window shows it in the ms-dos style:
> octave.exe:22> dir
> .              ..             lSmp=F7tila
> (an uppercase sigma in place of the a umlaut and a division symbol in
> place of the o umlaut).
>
> Octave doesn't react to hitting the national character keys on the
> keyboard. In order to chdir to that subdirectory I have to create an
> m-file containing the cd command. Proper handling of national
> characters in Octave might be a larger project, though. The data type
> used is signed 8-bit which cannot hold the necessary codes:
> octave.exe:23> double(char(246))
> ans =3D -10
> This should return 246 (the code for letter =F6).

This might be something to fix in the readline library I use (using differe=
nt
compilation switches). But, as you stated above, some problems might also
be tackled at octave-level (not MSVC-specific).

> 2. Windows Vista, relatively new setup without too many programs installe=
d
>
> The package installed without any problems. However, when trying to run,
> the Console window doesn't show any output from Octave. Opening a new
> Octave tab in Console shows an error message:
>
> error: unable to find Java Runtime Environment

Guess what? You'll need Java to run the java-based backend... :-)
More seriously, do you have JRE installed? OTOH, I should also make the
installer smarter and not install by default java-based backend if Java is
not detected on the target system.

Michael.

Re: [OctDev] Octave/Win32 new binary package (2.9.12)

[Olli S. <Oll...@kc...>]

>> error: unable to find Java Runtime Environment
> 
> Guess what? You'll need Java to run the java-based backend... :-)
> More seriously, do you have JRE installed? OTOH, I should also make the
> installer smarter and not install by default java-based backend if Java is
> not detected on the target system.

Or you might add a notification that tells the user to download and 
install JRE. An average user doesn't know whether the Octave package is 
supposed to contain all the necessary components. (I must be spoiled 
after using Windows for quite a few years :-) Anyway, it could be a 
barrier for several potential users if installing and starting Octave 
just pops up a blank window. Not even an error message without some 
further user action.

Another minor observation: Uninstall didn't remove start menu items on
Vista (even with JRE installed before installing & uninstalling Octave). 
On XP this worked as expected.

Olli

Re: [OctDev] Octave/Win32 new binary package (2.9.12)

[Michael G. <mic...@gm...>]

On 6/1/07, Olli Saarela <Oll...@kc...> wrote:
>
> >> error: unable to find Java Runtime Environment
> >
> > Guess what? You'll need Java to run the java-based backend... :-)
> > More seriously, do you have JRE installed? OTOH, I should also make the
> > installer smarter and not install by default java-based backend if Java is
> > not detected on the target system.
>
> Or you might add a notification that tells the user to download and
> install JRE. An average user doesn't know whether the Octave package is
> supposed to contain all the necessary components. (I must be spoiled
> after using Windows for quite a few years :-) Anyway, it could be a
> barrier for several potential users if installing and starting Octave
> just pops up a blank window. Not even an error message without some
> further user action.

I can indeed add such a warning.

> Another minor observation: Uninstall didn't remove start menu items on
> Vista (even with JRE installed before installing & uninstalling Octave).
> On XP this worked as expected.

This is something I really cannot test myself. I don't have Windows Vista.
Does it remove at least some entries, or none of them? Did the start menu
folder location change under Vista? Can you delete them manually?

Michael.

Re: [OctDev] Octave/Win32 new binary package (2.9.12)

[Olli S. <oll...@kc...>]

>> Another minor observation: Uninstall didn't remove start menu items on
>> Vista (even with JRE installed before installing & uninstalling Octave).
>> On XP this worked as expected.
> This is something I really cannot test myself. I don't have Windows 
> Vista.
> Does it remove at least some entries, or none of them? Did the start menu
> folder location change under Vista? Can you delete them manually?
Uninstall didn't remove any of the entries, but I was able to delete 
them manually by right-clicking the start menu.

I am currently not at my Vista box, I'll get back to you about the start 
folder location. I'll also be happy to run other tests that might help.

Olli

Re: [OctDev] Octave/Win32 new binary package (2.9.12)

[Michael G. <mic...@gm...>]

On 6/1/07, Olli Saarela <oll...@kc...> wrote:
> I am currently not at my Vista box, I'll get back to you about the start
> folder location. I'll also be happy to run other tests that might help.

Probably this weekend or early next week, a new package iteration
(let's call it a release candidate) should be available. At that time, some
new testing will be required. Thanks.

Michael.

[OctDev] File Handle leak? Was: Octave/Win32 new binary package (2.9.12)

[Benjamin L. <lin...@gm...>]

> Hi,
> 
> A new binary package of Octave for Win32, based on 2.9.12 is available on
> http://www.dbateman.org/?page=octave&lang=EN. Up to now, it's more a
> release candidate than an official package. I'd like to find a few people
> willing to give it a try and report problems, such that it can maybe
> become
> more official after fixing (it's based on official releases of octave
> and octave-forge).
> Anyone interested? Thanks.
> 
> Michael.

I tested the binary on a w2ksp4 machine and found that somehow it seems to
leak file handles.
Using ProcessExplorer the handle count steadily increases during usage.

A simple command like 'upper("a");' entered at the prompt increases 
the file handle count by 2 for every command.
The command 'plot(1:10,1:10);' using the oplot-gl package increases 
handle count by 48 for every call.

However wrapping command in a loop does not increase for every call.
E.g. 'for n=1:1000, b=upper("a"); endfor;' increases handle count by 2 like a single call.

Also calling built-in functions like sin(), cos() don't increase handle count.

Looking at the open handles I can see multiple handles to one and the same script file (e.g. upper.m, or drawnow.m)
This looks like octave creating a handle when reading/parsing a  script file and not releasing it.
I have tested with all possible settings of ignore_function_time_stamp and the behaviour does not change.

Can anyone confirm this ?

benjamin

P.S:
starting octave --no-init-file lists 3 file handles to share\octave\packages\java-1.0.1\octave.jar
executing at the prompt 'upper("a");' lists 2 additional handles to share\octave\2.9.12\m\strings\upper.m
every additional 'upper("a");' adds 2 file handles to share\octave\2.9.12\m\strings\upper.m


-- 
GMX FreeMail: 1 GB Postfach, 5 E-Mail-Adressen, 10 Free SMS.
Alle Infos und kostenlose Anmeldung: http://www.gmx.net/de/go/freemail

Re: [OctDev] File Handle leak? Was: Octave/Win32 new binary package (2.9.12)

[Olli S. <Oll...@kc...>]

Benjamin Lindner wrote:
> A simple command like 'upper("a");' entered at the prompt increases 
> the file handle count by 2 for every command.

The same happens on XP / Gnuplot.

> The command 'plot(1:10,1:10);' using the oplot-gl package increases 
> handle count by 48 for every call.

With Gnuplot on XP, a single 'plot(1:10,1:10);' command increases handle 
count by 60 in Octave and by a surprisingly high number 660 in pgnuplot.

Loop 'for j=1:10;plot(1:10,1:10);pause(1);end' increases handle count by 
60 in Octave and by 6600 in pgnuplot.

Command 'dir(".")' increases handle count by 16-22, for some reason the 
delta varies.

Olli

Re: [OctDev] File Handle leak? Was: Octave/Win32 new binary package (2.9.12)

[Michael G. <mic...@gm...>]

On 6/4/07, Olli Saarela <Oll...@kc...> wrote:
> Benjamin Lindner wrote:
> > A simple command like 'upper("a");' entered at the prompt increases
> > the file handle count by 2 for every command.

There was indeed a bug in octave code, where file handles were not closed
properly. See patch below.

Michael.

Index: sysdep.cc
===================================================================
RCS file: /cvs/octave/src/sysdep.cc,v
retrieving revision 1.128
diff -c -p -r1.128 sysdep.cc
*** sysdep.cc   31 May 2007 19:39:12 -0000      1.128
--- sysdep.cc   4 Jun 2007 14:59:45 -0000
*************** same_file_internal (const std::string& f
*** 269,274 ****
--- 269,277 ----
        CloseHandle (hfile2);
        return false;
      }
+
+   CloseHandle (hfile1);
+   CloseHandle (hfile2);

    return (hfi1.dwVolumeSerialNumber == hfi2.dwVolumeSerialNumber
          && hfi1.nFileIndexHigh == hfi2.nFileIndexHigh

Re: [OctDev] File Handle leak? Was: Octave/Win32 new binary package (2.9.12)

[Benjamin L. <lin...@gm...>]

> On 6/4/07, Olli Saarela <Oll...@kc...> wrote:
> > Benjamin Lindner wrote:
> > > A simple command like 'upper("a");' entered at the prompt increases
> > > the file handle count by 2 for every command.
> 
> There was indeed a bug in octave code, where file handles were not closed
> properly. See patch below.
> 
> Michael.

Ouch. I should have seen that.
Thanks for pointing out.
I checked, and the observed handle count increase indeed vanished (on a mingw32 build, this is, but it shares the same code here)

benjamin
-- 
GMX FreeMail: 1 GB Postfach, 5 E-Mail-Adressen, 10 Free SMS.
Alle Infos und kostenlose Anmeldung: http://www.gmx.net/de/go/freemail

Re: [OctDev] File Handle leak? Was: Octave/Win32 new binary package (2.9.12)

[Doug S. <da...@sy...>]

Are you going to post a new version with this fix, or should I download 
this on with the bug?

Doug

Michael Goffioul wrote:
> On 6/4/07, Olli Saarela <Oll...@kc...> wrote:
>   
>> Benjamin Lindner wrote:
>>     
>>> A simple command like 'upper("a");' entered at the prompt increases
>>> the file handle count by 2 for every command.
>>>       
>
> There was indeed a bug in octave code, where file handles were not closed
> properly. See patch below.
>
> Michael.
>
> Index: sysdep.cc
> ===================================================================
> RCS file: /cvs/octave/src/sysdep.cc,v
> retrieving revision 1.128
> diff -c -p -r1.128 sysdep.cc
> *** sysdep.cc   31 May 2007 19:39:12 -0000      1.128
> --- sysdep.cc   4 Jun 2007 14:59:45 -0000
> *************** same_file_internal (const std::string& f
> *** 269,274 ****
> --- 269,277 ----
>         CloseHandle (hfile2);
>         return false;
>       }
> +
> +   CloseHandle (hfile1);
> +   CloseHandle (hfile2);
>
>     return (hfi1.dwVolumeSerialNumber == hfi2.dwVolumeSerialNumber
>           && hfi1.nFileIndexHigh == hfi2.nFileIndexHigh
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by DB2 Express
> Download DB2 Express C - the FREE version of DB2 express and take
> control of your XML. No limits. Just data. Click to get it now.
> http://sourceforge.net/powerbar/db2/
> _______________________________________________
> Octave-dev mailing list
> Oct...@li...
> https://lists.sourceforge.net/lists/listinfo/octave-dev
>
>   

Re: [OctDev] File Handle leak? Was: Octave/Win32 new binary package (2.9.12)

[Michael G. <mic...@gm...>]

On 6/4/07, Doug Stewart <da...@sy...> wrote:
> Are you going to post a new version with this fix, or should I download
> this on with the bug?

>From a testing perspective, the new version is OK; but it's not OK for a
release as this bug will make octave rapidly eat up system resources.
To avoid making too much release candidates, what I'll probably do is let
people test this new version and fix potential problems along with this
file handle bug in the next release candidate (which is needed anyway).

Michael.

Re: [OctDev] File Handle leak? Was: Octave/Win32 new binary package (2.9.12)

[John W. E. <jw...@be...>]

On  4-Jun-2007, Michael Goffioul wrote:

| On 6/4/07, Olli Saarela <Oll...@kc...> wrote:
| > Benjamin Lindner wrote:
| > > A simple command like 'upper("a");' entered at the prompt increases
| > > the file handle count by 2 for every command.
| 
| There was indeed a bug in octave code, where file handles were not closed
| properly. See patch below.
| 
| Michael.
| 
| Index: sysdep.cc
| ===================================================================
| RCS file: /cvs/octave/src/sysdep.cc,v
| retrieving revision 1.128
| diff -c -p -r1.128 sysdep.cc
| *** sysdep.cc   31 May 2007 19:39:12 -0000      1.128
| --- sysdep.cc   4 Jun 2007 14:59:45 -0000
| *************** same_file_internal (const std::string& f
| *** 269,274 ****
| --- 269,277 ----
|         CloseHandle (hfile2);
|         return false;
|       }
| +
| +   CloseHandle (hfile1);
| +   CloseHandle (hfile2);
| 
|     return (hfi1.dwVolumeSerialNumber == hfi2.dwVolumeSerialNumber
|           && hfi1.nFileIndexHigh == hfi2.nFileIndexHigh

I rewrote it like this:

  bool retval = false;

  // Windows native code 
  // Reference: http://msdn2.microsoft.com/en-us/library/aa363788.aspx

  HANDLE hfile1 = CreateFile (file1.c_str (), 0, FILE_SHARE_READ, 0,
			      OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL, 0); 

  if (hfile1 != INVALID_HANDLE_VALUE)
    {
      HANDLE hfile2 = CreateFile (file2.c_str (), 0, FILE_SHARE_READ, 0,
				  OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL, 0);

      if (hfile2 != INVALID_HANDLE_VALUE)
	{  
	  BY_HANDLE_FILE_INFORMATION hfi1;
	  BY_HANDLE_FILE_INFORMATION hfi2;
  
	  if (GetFileInformationByHandle (hfile1, &hfi1)
	      && GetFileInformationByHandle (hfile2, &hfi2))
  
	    retval = (hfi1.dwVolumeSerialNumber == hfi2.dwVolumeSerialNumber
		      && hfi1.nFileIndexHigh == hfi2.nFileIndexHigh
		      && hfi1.nFileIndexLow == hfi2.nFileIndexLow);

	  CloseHandle (hfile2);
	}

      CloseHandle (hfile1);
    }

  return retval;


Is that OK?

jwe

Re: [OctDev] File Handle leak? Was: Octave/Win32 new binary package (2.9.12)

[Michael G. <mic...@gm...>]

On 6/5/07, John W. Eaton <jw...@be...> wrote:
> I rewrote it like this:
>
>  bool retval = false;
>
>  // Windows native code
>  // Reference: http://msdn2.microsoft.com/en-us/library/aa363788.aspx
>
>  HANDLE hfile1 = CreateFile (file1.c_str (), 0, FILE_SHARE_READ, 0,
>                              OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL, 0);
>
>  if (hfile1 != INVALID_HANDLE_VALUE)
>    {
>      HANDLE hfile2 = CreateFile (file2.c_str (), 0, FILE_SHARE_READ, 0,
>                                  OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL, 0);
>
>      if (hfile2 != INVALID_HANDLE_VALUE)
>        {
>          BY_HANDLE_FILE_INFORMATION hfi1;
>          BY_HANDLE_FILE_INFORMATION hfi2;
>
>          if (GetFileInformationByHandle (hfile1, &hfi1)
>              && GetFileInformationByHandle (hfile2, &hfi2))
>
>            retval = (hfi1.dwVolumeSerialNumber == hfi2.dwVolumeSerialNumber
>                      && hfi1.nFileIndexHigh == hfi2.nFileIndexHigh
>                      && hfi1.nFileIndexLow == hfi2.nFileIndexLow);
>
>          CloseHandle (hfile2);
>        }
>
>      CloseHandle (hfile1);
>    }
>
>  return retval;
>
>
> Is that OK?

Yes.

Michael.

Re: [OctDev] Octave/Win32 new binary package (2.9.12)

[Paul T. <pau...@wa...>]

Michael,

On starting up the installer, I get the slightly perplexing message:

GNU Octave 2.9.12 Setup

This version of Octave cannot be installed on this system.
Supported systems are Windows 2000 and Windows XP.

Current System: Microsoft Windows XP

This is the Professional X64 System

Cheers

Paul Thomas

Re: [OctDev] Octave/Win32 new binary package (2.9.12)

[Michael G. <mic...@gm...>]

On 6/11/07, Paul Thomas <pau...@wa...> wrote:
> Michael,
>
> On starting up the installer, I get the slightly perplexing message:
>
> GNU Octave 2.9.12 Setup
>
> This version of Octave cannot be installed on this system.
> Supported systems are Windows 2000 and Windows XP.
>
> Current System: Microsoft Windows XP
>
> This is the Professional X64 System

I added some code in the installer to detect the windows version; the main
reason was that Windows2000 requires a different installation of the runtime
libs than WinXP (and Vista). Recognized windows versions are 5.0 (Win2K),
5.1 (WinXP) and 6.x (Vista). What's the content of the registry key
HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows NT\CurrentVersion?

Michael.