octave-dev

[OctDev] Re: A new standalone octave for 2.1.60, and setup.

[Paul K. <pki...@us...>]

Ole,

Here's an outline of what needs to be done for a new Windows release.

I'm CC'ing Brian Bliss and the octave-forge list in case anyone wants
to help you.

I've put my compilation of octave-2.1.60.tar.bz2 into the root
of waterthrill.  Expand it in the /opt directory.  I get a
"Permission denied" error when I quit octave.  Let me know if
it works better for you.

The symbolic linking stuff isn't finished yet.  I didn't realize
earlier that octave was doing hard links.  Cygwin soft links
might install automatically, but it may be that link implementation
might be windows version specific so we may need to force the link
during install.  I've updated the link scripts generated by
make to remove the old symlink before installing the new.  I'm attaching
mklinks.sh for converting the /opt/octave-2.1.60/base/oct via

	$ source mklinks.sh

when in that directory.  This should be run from the current
build script but it isn't. The current post-install script runs
mklinks again in the right directories, but this is untested.

I know the octave.nsi NSIS script is broken.  I've been hacking at
it without testing it.  I thought it was better to post and not
lose what I was working on since we didn't really have a
working build anyway.

You will need to do some work to make sure the proper directories
are installed. cygwin, bin, base and site should do it.  E.g.,

	SetOutPath $INSTDIR\bin
     FILE /r "${ROOT}\opt\octave-${VERSION}\bin\*.*"

Next you will have to set the Cygwin registry keys so that the
appropriate directories are mounted.  E.g.,

	/bin -> $INSTDIR/cygwin
     /usr/bin -> $INSTDIR/bin
	/opt/octave-${VERSION} -> $INSTDIR
	/ -> $INSTDIR
	/home -> $DOCUMENTS

I'm not entirely convinced this will work.  Can you have /usr/bin
without having /usr?  Can you have / and /opt/octave-2.1.60 pointing
to the same directory?

Make sure to create directories for tmp and for the users octave
files:

	CreateDirectory $INSTDIR\tmp.
	CreateDirectory $DOCUMENTS\octave

Make sure that CreateDirectory doesn't fail if the directory
already exists, or empty it if it does --- we wouldn't want to
delete the user's private octave scripts.  Presumably the HOME
environment variable should be set to /home/octave.  Somewhere
there is a system startup file which should automatically include
/home/octave on the octave path.

Eliminate as much as you can from the octave_install.sh script.
I bet NSIS is capable of doing everything we need.  Particularly,
setting up an editor and predefining all the constants for edit.m
would be nice.

Figure out what's up with the o|c|t|a|v|e|>| prompt which
people sometimes see.

We should be able to do a drop-in replacement with the atlas dlls.
I haven't constructed the atlas dlls so I've never tried this.
We should make sure we don't overwrite lapack.dll when we do the
replacement.

Be sure to test both system install and user install.

I would like a nag screen asking for Octave donations. 30000
downloads x $10/year is 3 full time employees working on octave.

I'm not going to discuss the requirements for a dev release here.

This should be enough to get you started, but I'm sure there are
other details.  There is a reason I haven't done a release in a
while 8-)  Once we work out the kinks in this release, it will be
smooth sailing in the future.

- Paul

On Oct 27, 2004, at 6:11 PM, Ole Jacob Hagen wrote:

> You can send me your standalone there...
> Which version of NSIS are you using? I am using the new release....It 
> doesn't work properly...Have to edit the nsi-file...
>

Re: [OctDev] Re: A new standalone octave for 2.1.60, and setup.

[Ole J. H. <wat...@ya...>]

Hi, Paul.

During the late hours tonight I managed to make a standalone 
Octave-2.1.60 with Octave-forge and win32-version of Gnuplot.

Now I've just have to solve the symlinks of the oct-files, and go 
through all the oct-files.

I didn't get a "permission denied" error when I quit Octave.

I've installed octave-forge into Octave's /share and libexec as done on 
Linux-systems.
Why this doesn't happen automatically in cygwin environment, I don't know.
But it should, right?

Thanks for helping out, Paul.
Appreciate it a lot.

This didn't work on the new version of NSIS (2.0):

    FILE /r "${ROOT}\opt\octave-${VERSION}\bin\*.*"

I should have a standalone application this day.

Need a suggestion:

Should I use ATLAS and/or BLAS packages?

Cheers,

Ole J.

Re: [OctDev] Re: A new standalone octave for 2.1.60, and setup.

[Paul K. <pki...@us...>]

So you rebuilt from scratch?  Or did you use my version?
Obviously the former since my version had octave-forge
already installed in /opt/octave-2.1.60/site/octave-forge.

Did you use octave.nsi?  What changes did you make?

What did you use for the cygwin environment?  The one I
provided?  Have you tested it yet?  Did it work?

My lapack.dll includes blas.  The idea was to make
platform specific lapack-atlas-arch.dll files and copy
them over lapack.dll at install time.  Note that we
should have a copy of lapack.dll in lapack-base.dll
in case there is any problems.

I don't know why the FILE command isn't working.

- Paul

On Oct 28, 2004, at 2:31 AM, Ole Jacob Hagen wrote:

> Hi, Paul.
>
> During the late hours tonight I managed to make a standalone 
> Octave-2.1.60 with Octave-forge and win32-version of Gnuplot.
>
> Now I've just have to solve the symlinks of the oct-files, and go 
> through all the oct-files.
>
> I didn't get a "permission denied" error when I quit Octave.
>
> I've installed octave-forge into Octave's /share and libexec as done 
> on Linux-systems.
> Why this doesn't happen automatically in cygwin environment, I don't 
> know.
> But it should, right?
>
> Thanks for helping out, Paul.
> Appreciate it a lot.
>
> This didn't work on the new version of NSIS (2.0):
>
>    FILE /r "${ROOT}\opt\octave-${VERSION}\bin\*.*"
>
> I should have a standalone application this day.
>
> Need a suggestion:
>
> Should I use ATLAS and/or BLAS packages?
>
> Cheers,
>
> Ole J.
>

Re: [OctDev] Re: A new standalone octave for 2.1.60, and setup.

[Ole J. H. <wat...@ya...>]

Hi, Paul.

Yes, I used octave.nsi, and No, I didn't use the file you provided.
I've built Octave, octave-forge from scratch.

I've tested the gshow function, and my gnuplot pops up. ;-)

I can install Octave by using the installer, which works.
I've tried NSIS-2.02 and 2.0, and they both are giving this error:
What version of NSIS did you make the Octave-2.1.50 package with?

This is what doesn't work(, and what to do... )

No Language selection appears: This is easy to solve.
No License text appears: It shouldn't be difficult to fix.

At the end of the installation: Windows can't find run.exe, and throws 
this error message at me:

"Error: Couldn't find Files\GNU anywhere.
I even looked in the PATH, I also tried appending
[1]: 
[2]:.exe
"

Seen this error before?

And rxvt doesn't pop up either...What libraries are rxvt depending on? 
I've tried to compare with the install-package for Octave-2.1.50.

I've compared the entries in REGEDIT as well. Same entries are found in 
both Octave-2.1.50 and Octave-2.1.60.
I believe the problem isn't here....Rather the PATH settings, I guess?

Personally I can survive without the language and license, but when 
install_octave.sh is called I'm getting the error mentioned above.

Any suggestions?

Ole



Paul Kienzle wrote:

> So you rebuilt from scratch?  Or did you use my version?
> Obviously the former since my version had octave-forge
> already installed in /opt/octave-2.1.60/site/octave-forge.
>
> Did you use octave.nsi?  What changes did you make?
>
> What did you use for the cygwin environment?  The one I
> provided?  Have you tested it yet?  Did it work?
>
> My lapack.dll includes blas.  The idea was to make
> platform specific lapack-atlas-arch.dll files and copy
> them over lapack.dll at install time.  Note that we
> should have a copy of lapack.dll in lapack-base.dll
> in case there is any problems.
>
> I don't know why the FILE command isn't working.
>
> - Paul
>
> On Oct 28, 2004, at 2:31 AM, Ole Jacob Hagen wrote:
>
>> Hi, Paul.
>>
>> During the late hours tonight I managed to make a standalone 
>> Octave-2.1.60 with Octave-forge and win32-version of Gnuplot.
>>
>> Now I've just have to solve the symlinks of the oct-files, and go 
>> through all the oct-files.
>>
>> I didn't get a "permission denied" error when I quit Octave.
>>
>> I've installed octave-forge into Octave's /share and libexec as done 
>> on Linux-systems.
>> Why this doesn't happen automatically in cygwin environment, I don't 
>> know.
>> But it should, right?
>>
>> Thanks for helping out, Paul.
>> Appreciate it a lot.
>>
>> This didn't work on the new version of NSIS (2.0):
>>
>>    FILE /r "${ROOT}\opt\octave-${VERSION}\bin\*.*"
>>
>> I should have a standalone application this day.
>>
>> Need a suggestion:
>>
>> Should I use ATLAS and/or BLAS packages?
>>
>> Cheers,
>>
>> Ole J.
>>
>
>

Re: [OctDev] Re: A new standalone octave for 2.1.60, and setup.

[Ole J. H. <ol...@wa...>]

Hi,

Yippi kay ay, I am making progress. ;-)

I've solved the Registry entries now.
There were two different definitions of OCTKEY in octave.nsi and  
Makefile, which made two  registry entries for our stand alone octave.

Now I need to fix the symlinks in my oct-path's.
The function dispatch is called by default, isn't it?
During startup of Octave I'm getting this warning message:
Warning: potential Matlab compatiblity problem: " used as string delimiter.

Where is the origin of this warning? What file? Is it octave-core of 
sf-octave?

When I'm calling help <function> I am getting this error:
error: feval: the symbol 'dispatch_help'  is not valid as a function.

I guess this is because of the symlinks? Maybe it's a hardlink?

Have you made a script that I can use to change the hardlinks into 
symlinks again?
That would be terrific if you have made such script? Or maybe there is 
someone who has made it before me?

I could always remove octave-forge from my octave distribution, but that 
would be very unfortunate. ;-)

Cheers,

Ole

Re: [OctDev] Re: A new standalone octave for 2.1.60, and setup.

[Ole J. H. <wat...@ya...>]

Hi, again.

Do you have some suggestions how to override or remove the warnings I am 
getting?
That would be terrific.


Cheers,

Ole J.

A benchmark test,  with my Octave, compiled with gcc-3.3.3...
Fails in the rest, since we have no gcd2....can I use gcd?

>   I. Matrix calculation
>    ---------------------
>
> Creation, transp., deformation of a 1500x1500 matrix (sec): 1.008
> 800x800 normal distributed random matrix ^1000______ (sec): 0.8307
> Sorting of 2,000,000 random values__________________ (sec): 1.291
> 700x700 cross-product matrix (b = a' * a)___________ (sec): 4.181
> Linear regression over a 600x600 matrix (c = a \ b') (sec): 0.9707
>                   ------------------------------------------------------
> FFT over 800,000 random values______________________ (sec): 0.9987
> Eigenvalues of a 320x320 random matrix______________ (sec): 1.196
> Determinant of a 650x650 random matrix______________ (sec): 1.216
> Cholesky decomposition of a 900x900 matrix__________ (sec): 0.6573
> Inverse of a 400x400 random matrix__________________ (sec): 0.8017
>                   ------------------------------------------------------
>                 Trimmed geom. mean (2 extremes eliminated): 0.9855
>  
>    III. Programmation
>    ------------------
> 750,000 Fibonacci numbers calculation (vector calc)_ (sec): 1.772
> Creation of a 2250x2250 Hilbert matrix (matrix calc) (sec): 1.078
>

Re: [OctDev] Re: A new standalone octave for 2.1.60, and setup.

[David B. <Dav...@mo...>]

gcd2.m was delivered with this benchmark. I attach a copy here.

D.

According to Ole Jacob Hagen <wat...@ya...> (on 10/29/04):
> Hi, again.
> 
> Do you have some suggestions how to override or remove the warnings I am 
> getting?
> That would be terrific.
> 
> 
> Cheers,
> 
> Ole J.
> 
> A benchmark test,  with my Octave, compiled with gcc-3.3.3...
> Fails in the rest, since we have no gcd2....can I use gcd?
> 
> >  I. Matrix calculation
> >   ---------------------
> >
> >Creation, transp., deformation of a 1500x1500 matrix (sec): 1.008
> >800x800 normal distributed random matrix ^1000______ (sec): 0.8307
> >Sorting of 2,000,000 random values__________________ (sec): 1.291
> >700x700 cross-product matrix (b = a' * a)___________ (sec): 4.181
> >Linear regression over a 600x600 matrix (c = a \ b') (sec): 0.9707
> >                  ------------------------------------------------------
> >FFT over 800,000 random values______________________ (sec): 0.9987
> >Eigenvalues of a 320x320 random matrix______________ (sec): 1.196
> >Determinant of a 650x650 random matrix______________ (sec): 1.216
> >Cholesky decomposition of a 900x900 matrix__________ (sec): 0.6573
> >Inverse of a 400x400 random matrix__________________ (sec): 0.8017
> >                  ------------------------------------------------------
> >                Trimmed geom. mean (2 extremes eliminated): 0.9855
> > 
> >   III. Programmation
> >   ------------------
> >750,000 Fibonacci numbers calculation (vector calc)_ (sec): 1.772
> >Creation of a 2250x2250 Hilbert matrix (matrix calc) (sec): 1.078
> >
> 
> 
> 
> 
> -------------------------------------------------------
> This Newsletter Sponsored by: Macrovision 
> For reliable Linux application installations, use the industry's leading
> setup authoring tool, InstallShield X. Learn more and evaluate 
> today. http://clk.atdmt.com/MSI/go/ins0030000001msi/direct/01/
> _______________________________________________
> Octave-dev mailing list
> Oct...@li...
> https://lists.sourceforge.net/lists/listinfo/octave-dev

-- 
David Bateman                                Dav...@mo...
Motorola CRM                                 +33 1 69 35 48 04 (Ph) 
Parc Les Algorithmes, Commune de St Aubin    +33 1 69 35 77 01 (Fax) 
91193 Gif-Sur-Yvette FRANCE

The information contained in this communication has been classified as: 

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

Re: [OctDev] Re: A new standalone octave for 2.1.60, and setup.

[Ole J. H. <wat...@ya...>]

Hi,

Yippi kay ay, I am making progress. ;-)

I've solved the Registry entries now.
There were two different definitions of OCTKEY in octave.nsi and
Makefile, which made two  registry entries for our stand alone octave.

Now I need to fix the symlinks in my oct-path's.
The function dispatch is called by default, isn't it?
During startup of Octave I'm getting this warning message:
Warning: potential Matlab compatiblity problem: " used as string delimiter.

Where is the origin of this warning? What file? Is it octave-core of
sf-octave?

When I'm calling help <function> I am getting this error:
error: feval: the symbol 'dispatch_help'  is not valid as a function.

I guess this is because of the symlinks? Maybe it's a hardlink?

Have you made a script that I can use to change the hardlinks into
symlinks again?
That would be terrific if you have made such script? Or maybe there is
someone who has made it before me?

I could always remove octave-forge from my octave distribution, but that
would be very unfortunate. ;-)

Cheers,

Ole