octave-dev

[OctDev] Changes for system function

[Quentin S. <qsp...@ie...>]

A while ago I reported an incompatibility with Matlab in the order of 
outputs of the "system" function. John made changes in CVS yesterday to 
fix this. I looked through octave-forge CVS and found several instances 
where "system" is used that will also need to be changed. Should I check 
these changes in now, or wait until a new octave is released? I'm 
assuming we're now breaking compatibility with 2.1.x.

-Quentin

Re: [OctDev] Changes for system function

[Etienne G. <et...@cs...>]

On Fri, Mar 10, 2006 at 12:20:20PM -0600, Quentin Spencer wrote:
# A while ago I reported an incompatibility with Matlab in the order of 
# outputs of the "system" function. John made changes in CVS yesterday to 
# fix this. I looked through octave-forge CVS and found several instances 
# where "system" is used that will also need to be changed. Should I check 
# these changes in now, or wait until a new octave is released? I'm 
# assuming we're now breaking compatibility with 2.1.x.
# 
# -Quentin
[snip]


Iigc, this is both the two-output system call, for which
[msg,status]=system should become [status,msg]=system and the
single-uutput, for which msg=system should become
[dummyname,msg]=system.

  Seems feasible semi-automatically.

  The cases I find in OF are the following. Do you find the same?

======================================================================
./extra/Windows/dos.m:    [text, status] = system(['cmd /c ',cmd]);
./main/audio/sound.m:      host = system("uname -n");
./main/comm/comms.m:          showfile = system(["find '", infopath(1:len-1), "' -name ", ...
./main/comm/comms.m:          showfile = system(["find '", infopath, "' -name ", infofile, ...
./main/comm/comms.m:      [testout, testret] = system(["'", INFO_PROGRAM, "' --version"]);
./main/comm/comms.m:      [testout, testret] = system("info --version");
./main/fixed/fixedpoint.m:          showfile = system(["find '", infopath(1:len-1), "' -name ", ...
./main/fixed/fixedpoint.m:          showfile = system(["find '", infopath, "' -name ", infofile, ...
./main/fixed/fixedpoint.m:      [testout, testret] = system(["'", INFO_PROGRAM, "' --version"]);
./main/fixed/fixedpoint.m:      [testout, testret] = system("info --version");
./main/general/unix.m:    [text, status] = system (cmd);
./main/image/imginfo.m:[res,status] = system(sprintf("convert -verbose '%s' /dev/null",fn),1);
./main/image/imwrite.m:      [jnk,retcode] = system(cmd);
./main/image/myimage.m:    [o,s] = system ("feh tmp.pgm");
./main/image/myimage.m:    [o,s] = system ("feh tmp.ppm");
./main/miscellaneous/edit.m:      host = system("uname -n");
./main/miscellaneous/slurp_file.m:s = system (sprintf ("cat '%s'",f), 1);
./main/plot/print.m:    [output, errcode] = system (command);
./main/vrml/select_3D_points.m:res = system (cmd, 1);
./main/vrml/vrml_browse.m:  [dum,status]=system (sprintf ("kill -CONT %d &> /dev/null",vrml_b_pid));
./main/vrml/vrml_browse.m:[dum,status] = system (["test -e ",b_temp]);
./main/vrml/vrml_browse.m:  [out, status] = system (cmd = [vrml_b_name," ",b_opt," ",b_temp], 1);
./main/vrml/vrml_browse.m:  [dum,status]=system (sprintf ("kill -USR1 %d &> /dev/null",vrml_b_pid));
======================================================================

-- 
Etienne Grossmann ------ http://www.cs.uky.edu/~etienne

Re: [OctDev] Changes for system function

[Quentin S. <qsp...@ie...>]

Etienne Grossmann wrote:

>On Fri, Mar 10, 2006 at 12:20:20PM -0600, Quentin Spencer wrote:
># A while ago I reported an incompatibility with Matlab in the order of 
># outputs of the "system" function. John made changes in CVS yesterday to 
># fix this. I looked through octave-forge CVS and found several instances 
># where "system" is used that will also need to be changed. Should I check 
># these changes in now, or wait until a new octave is released? I'm 
># assuming we're now breaking compatibility with 2.1.x.
># 
># -Quentin
>[snip]
>
>
>Iigc, this is both the two-output system call, for which
>[msg,status]=system should become [status,msg]=system and the
>single-uutput, for which msg=system should become
>[dummyname,msg]=system.
>
>  Seems feasible semi-automatically.
>
>  The cases I find in OF are the following. Do you find the same?
>
>======================================================================
>./extra/Windows/dos.m:    [text, status] = system(['cmd /c ',cmd]);
>./main/audio/sound.m:      host = system("uname -n");
>./main/comm/comms.m:          showfile = system(["find '", infopath(1:len-1), "' -name ", ...
>./main/comm/comms.m:          showfile = system(["find '", infopath, "' -name ", infofile, ...
>./main/comm/comms.m:      [testout, testret] = system(["'", INFO_PROGRAM, "' --version"]);
>./main/comm/comms.m:      [testout, testret] = system("info --version");
>./main/fixed/fixedpoint.m:          showfile = system(["find '", infopath(1:len-1), "' -name ", ...
>./main/fixed/fixedpoint.m:          showfile = system(["find '", infopath, "' -name ", infofile, ...
>./main/fixed/fixedpoint.m:      [testout, testret] = system(["'", INFO_PROGRAM, "' --version"]);
>./main/fixed/fixedpoint.m:      [testout, testret] = system("info --version");
>./main/general/unix.m:    [text, status] = system (cmd);
>./main/image/imginfo.m:[res,status] = system(sprintf("convert -verbose '%s' /dev/null",fn),1);
>./main/image/imwrite.m:      [jnk,retcode] = system(cmd);
>./main/image/myimage.m:    [o,s] = system ("feh tmp.pgm");
>./main/image/myimage.m:    [o,s] = system ("feh tmp.ppm");
>./main/miscellaneous/edit.m:      host = system("uname -n");
>./main/miscellaneous/slurp_file.m:s = system (sprintf ("cat '%s'",f), 1);
>./main/plot/print.m:    [output, errcode] = system (command);
>./main/vrml/select_3D_points.m:res = system (cmd, 1);
>./main/vrml/vrml_browse.m:  [dum,status]=system (sprintf ("kill -CONT %d &> /dev/null",vrml_b_pid));
>./main/vrml/vrml_browse.m:[dum,status] = system (["test -e ",b_temp]);
>./main/vrml/vrml_browse.m:  [out, status] = system (cmd = [vrml_b_name," ",b_opt," ",b_temp], 1);
>./main/vrml/vrml_browse.m:  [dum,status]=system (sprintf ("kill -USR1 %d &> /dev/null",vrml_b_pid));
>======================================================================
>
>  
>

This list is pretty close to mine, although I found a couple of others. 
I'll check in the changes soon, but we'll need to be aware of this after 
2.9.5 is released and make sure there isn't anything I missed.

-Quentin

Re: [OctDev] Changes for system function

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

Quentin Spencer wrote:
> Etienne Grossmann wrote:
>
>> On Fri, Mar 10, 2006 at 12:20:20PM -0600, Quentin Spencer wrote:
>> # A while ago I reported an incompatibility with Matlab in the order 
>> of # outputs of the "system" function. John made changes in CVS 
>> yesterday to # fix this. I looked through octave-forge CVS and found 
>> several instances # where "system" is used that will also need to be 
>> changed. Should I check # these changes in now, or wait until a new 
>> octave is released? I'm # assuming we're now breaking compatibility 
>> with 2.1.x.
>> # # -Quentin
>> [snip]
>>
>>
>> Iigc, this is both the two-output system call, for which
>> [msg,status]=system should become [status,msg]=system and the
>> single-uutput, for which msg=system should become
>> [dummyname,msg]=system.
>>
>>  Seems feasible semi-automatically.
>>
>>  The cases I find in OF are the following. Do you find the same?
>>
>> ======================================================================
>> ./extra/Windows/dos.m:    [text, status] = system(['cmd /c ',cmd]);
>> ./main/audio/sound.m:      host = system("uname -n");
>> ./main/comm/comms.m:          showfile = system(["find '", 
>> infopath(1:len-1), "' -name ", ...
>> ./main/comm/comms.m:          showfile = system(["find '", infopath, 
>> "' -name ", infofile, ...
>> ./main/comm/comms.m:      [testout, testret] = system(["'", 
>> INFO_PROGRAM, "' --version"]);
>> ./main/comm/comms.m:      [testout, testret] = system("info --version");
>> ./main/fixed/fixedpoint.m:          showfile = system(["find '", 
>> infopath(1:len-1), "' -name ", ...
>> ./main/fixed/fixedpoint.m:          showfile = system(["find '", 
>> infopath, "' -name ", infofile, ...
>> ./main/fixed/fixedpoint.m:      [testout, testret] = system(["'", 
>> INFO_PROGRAM, "' --version"]);
>> ./main/fixed/fixedpoint.m:      [testout, testret] = system("info 
>> --version");
>> ./main/general/unix.m:    [text, status] = system (cmd);
>> ./main/image/imginfo.m:[res,status] = system(sprintf("convert 
>> -verbose '%s' /dev/null",fn),1);
>> ./main/image/imwrite.m:      [jnk,retcode] = system(cmd);
>> ./main/image/myimage.m:    [o,s] = system ("feh tmp.pgm");
>> ./main/image/myimage.m:    [o,s] = system ("feh tmp.ppm");
>> ./main/miscellaneous/edit.m:      host = system("uname -n");
>> ./main/miscellaneous/slurp_file.m:s = system (sprintf ("cat '%s'",f), 
>> 1);
>> ./main/plot/print.m:    [output, errcode] = system (command);
>> ./main/vrml/select_3D_points.m:res = system (cmd, 1);
>> ./main/vrml/vrml_browse.m:  [dum,status]=system (sprintf ("kill -CONT 
>> %d &> /dev/null",vrml_b_pid));
>> ./main/vrml/vrml_browse.m:[dum,status] = system (["test -e ",b_temp]);
>> ./main/vrml/vrml_browse.m:  [out, status] = system (cmd = 
>> [vrml_b_name," ",b_opt," ",b_temp], 1);
>> ./main/vrml/vrml_browse.m:  [dum,status]=system (sprintf ("kill -USR1 
>> %d &> /dev/null",vrml_b_pid));
>> ======================================================================
>>
>>  
>>
>
> This list is pretty close to mine, although I found a couple of 
> others. I'll check in the changes soon, but we'll need to be aware of 
> this after 2.9.5 is released and make sure there isn't anything I missed.
>
> -Quentin
If this is purely a 2.9.5 change then I'd suggest that the comm and 
fixedpoint system calls that call info might be replaced with the doc.m 
function, and so it might be better to eliminate these than convert them..

D.

[OctDev] 2.1.x branch

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

I ran the following cvs command on the R2006-03-17 release:

	cvs tag -b R2-1-x

That means we now have a branch for 2.1.x.  Any fixes against the 2.1.x 
octave can be applied to there.

I used the following command to check out the branch:

	cvs -d :ext:pki...@cv...:/cvsroot/octave co 
-doctave-forge-21 -r R2-1-x octave-forge

I then modified the Makefile to note that it was in the 2.1.x branch, 
and indeed the changes did not get copied to the trunk.

I have no intention of making any more 2.1.x releases, but should some 
be needed, release.sh could be modified to accept a tag parameter when 
downloading the archive.

All pre-2.9.5 cruft can now be purged.

I would like to see a new 'octave-forge/package' tree.  Directories in 
this tree should install using the octave-2.9.x packaging system.  
Packages in main, extra and non-free should eventually move to package. 
  We should not need any toplevel configure script---if the octave 
packaging system doesn't do what we need it to, then extend the 
packaging system. octave-forge is dead!  Long live octave-forge.

Given the problems with cvs on sourceforge lately, we might consider 
moving to e.g., savannah and subversion.  At a minimum we could move to 
subversion on sourceforge.  The subversion page on sourceforge says 
that it is slower than CVS, but I understand that it will feel faster 
because more operations are done locally.  The main argument for moving 
to subversion, though, is that we can move packages from main to 
package without losing history.

Anyone care to do the work, and tell the rest of us what we need to do 
that is different?

- Paul

On Mar 10, 2006, at 1:20 PM, Quentin Spencer wrote:

> A while ago I reported an incompatibility with Matlab in the order of 
> outputs of the "system" function. John made changes in CVS yesterday 
> to fix this. I looked through octave-forge CVS and found several 
> instances where "system" is used that will also need to be changed. 
> Should I check these changes in now, or wait until a new octave is 
> released? I'm assuming we're now breaking compatibility with 2.1.x.
>
> -Quentin