[octave-forge] (parallel) request help for testing at Windows

classic Classic list List threaded Threaded
20 messages Options
Reply | Threaded
Open this post in threaded view
|

[octave-forge] (parallel) request help for testing at Windows

Olaf Till-2
Hi,

while re-structuring the parallel package to not work with forked
children of Octave sessions, I tried to make the local and client part
of the package compatible with Windows. (The server part will not be
built at windows.)

Currently the infrastructure for the new code is in place and
everything which is critical for Windows compatibility should be
done. Preliminary tests pass at current Octave versions at
GNU/Linux. The package cross-compiled successfully for mingw at
mxe-octave.

Could somebody run the tests described below at Windows? (Tests at
other non-GNU OSs are welcome, too, of course.) Since building at
Windows may need a package with pre-built docs, I uploaded one at:

https://sourceforge.net/p/octave/package-releases/393/

md5sum:
17bb3531634d6d40d0c09930d1c01165  parallel-3.1.3.tar.gz

It is from changeset a7efbe08499a.

Warning: Use a new Octave session and close it afterwards, i.e don't
use the same Octave session for something you need. Apart from
unforeseen crashes, the preliminary test function isn't crash-proof if
given the wrong arguments.

Requested test (for local parallel execution):

    'pkg load' the package and:

    test ("path\to\sourcefile\__test_rfeval__.cc")

Thanks in advance, if you come so far.

Olaf


PS:

For testing remote execution, the effort is by far greater. If someone
should nevertheless be willing to undertake it: The package must be
built and installed at two different machines, one with GNU/Linux, for
the server, and one with Windows (or another non-GNU OS), to test the
client functions.

    If you can't exclude unauthorized access to the server at ports
    12501 and 12502 (e.g. by a firewall, maybe in a separate router),
    you should only test with authenticated connections (really!).

    Authentication keys can be generated from the Octave prompt with
    'parallel_generate_srp_data ("arbitrary_user_name")', hitting
    ENTER when requested for a password. The resulting files should be
    placed at the server as described in the helptext of
    'parallel_generate_srp_data'.

    At the server machine, make sure that 'pkg load parallel' is
    called automatically at Octave startup. Then from the shell
    (substitute PREFIX twice):

    /PREFIX/share/octave/packages/parallel-3.1.3/bin/octave-pserver --octave-binary=/PREFIX/bin/octave-cli

    You may have to hit ENTER to see the next shell prompt. There is
    not yet away to kill the server except from the shell with 'kill
    <pid>'. You can get the pid with 'ps ax | grep pserver', e.g..

    At the client machine (Windows), start Octave, 'pkg load' the
    parallel package, and test the following (note the "{" and "}"):

    conn = pconnect ({"name.or.ip-address.of.servermachine"})
    assert (rfeval (@ (x) x, [3, 4, 5], 1, [], conn(2)), {[3, 4, 5]})
    sclose (conn)

    Kill the 'octave-pserver' at the server machine.

    If you are able to also test unauthorized connections (see warning
    above), at the server machine, after killing any running
    'octave-pserver', start a new one as described above, but with the
    additional option '--no-auth', i.e., from the shell:

    /PREFIX/share/octave/packages/parallel-3.1.3/bin/octave-pserver --octave-binary=/PREFIX/bin/octave-cli --no-auth

    At the client machine, the command for connecting is now:

    conn = pconnect ({"name.or.ip-address.of.servermachine"}, struct ("use_tls", false))

    The rest is the same as with authentication.

    Olaf

--
public key id EAFE0591, e.g. on x-hkp://pool.sks-keyservers.net

signature.asc (849 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: [octave-forge] (parallel) request help for testing at Windows

nrjank
got a build/install failure on windows with error messages that
overran the interpreter window buffer. took a while, so rerunning now
with -verbose and diary turned on.  will reply with diary attached when
it finishes.

Reply | Threaded
Open this post in threaded view
|

Re: [octave-forge] (parallel) request help for testing at Windows

nrjank
On Sat, Nov 23, 2019 at 1:03 PM Nicholas Jankowski <[hidden email]> wrote:
>
> got a build/install failure on windows with error messages that
> overran the interpreter window buffer. took a while, so rerunning now
> with -verbose and diary turned on.  will reply with diary attached when
> it finishes.

Huh.  so it seems diary doesn't capture pkg messages.  Bug??


diary contents:
--------------------
octave:5> pkg install -verbose parallel-3.1.3.tar.gz
mkdir (c:\Programs\Octave\Temp\oct-YQjsJD)
untar (parallel-3.1.3.tar.gz, c:\Programs\Octave\Temp\oct-YQjsJD)

error: called from
    install at line 192 column 5
    pkg at line 441 column 9
pkg: error running `make' for the parallel package.
error: called from
    configure_make at line 99 column 9
    install at line 184 column 7
    pkg at line 441 column 9
octave:5> diary off
---------------------

Reply | Threaded
Open this post in threaded view
|

Re: [octave-forge] (parallel) request help for testing at Windows

nrjank
On Sat, Nov 23, 2019 at 1:08 PM Nicholas Jankowski <[hidden email]> wrote:
>
> On Sat, Nov 23, 2019 at 1:03 PM Nicholas Jankowski <[hidden email]> wrote:
> >
> > got a build/install failure on windows with error messages that
> > overran the interpreter window buffer. took a while, so rerunning now
> > with -verbose and diary turned on.  will reply with diary attached when
> > it finishes.
>

ran again with console buffer set higher.  here you go. output saved
into attached file.

diary.txt (108K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: [octave-forge] (parallel) request help for testing at Windows

Olaf Till-2
On Sat, Nov 23, 2019 at 01:10:00PM -0500, Nicholas Jankowski wrote:

> On Sat, Nov 23, 2019 at 1:08 PM Nicholas Jankowski <[hidden email]> wrote:
> >
> > On Sat, Nov 23, 2019 at 1:03 PM Nicholas Jankowski <[hidden email]> wrote:
> > >
> > > got a build/install failure on windows with error messages that
> > > overran the interpreter window buffer. took a while, so rerunning now
> > > with -verbose and diary turned on.  will reply with diary attached when
> > > it finishes.
> >
>
> ran again with console buffer set higher.  here you go. output saved
> into attached file.
Thanks a lot!

> octave:5> pkg install -verbose parallel-3.1.3.tar.gz
> ...
> checking for gcc... gcc
> ...
> cc -c gnulib-wrappers.c -Igl -g -O2 -fPIC
> make: cc: Command not found

So configure finds gcc, but make uses 'cc'. I probably shouldn't make
CC configurable from environment in src/Makefile.in.

After fixing this (changeset 723cda56ca37), a new package with
pre-built docs has been uploaded, as before to:
https://sourceforge.net/p/octave/package-releases/393/
md5sum:
1fedf1ea96f8e249cd48421ef6a54a88  parallel-3.1.3.tar.gz

Olaf

--
public key id EAFE0591, e.g. on x-hkp://pool.sks-keyservers.net

signature.asc (849 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: [octave-forge] (parallel) request help for testing at Windows

nrjank
In reply to this post by nrjank

Huh.  so it seems diary doesn't capture pkg messages.  Bug??

added a comment to bug 54898 regarding diary capture

Reply | Threaded
Open this post in threaded view
|

Re: [octave-forge] (parallel) request help for testing at Windows

nrjank
In reply to this post by Olaf Till-2
On Sat, Nov 23, 2019 at 3:53 PM Olaf Till <[hidden email]> wrote:
On Sat, Nov 23, 2019 at 01:10:00PM -0500, Nicholas Jankowski wrote:
> On Sat, Nov 23, 2019 at 1:08 PM Nicholas Jankowski <[hidden email]> wrote:
> >
> > On Sat, Nov 23, 2019 at 1:03 PM Nicholas Jankowski <[hidden email]> wrote:
> > >
> > > got a build/install failure on windows with error messages that
> > > overran the interpreter window buffer. took a while, so rerunning now
> > > with -verbose and diary turned on.  will reply with diary attached when
> > > it finishes.
> >
>
> ran again with console buffer set higher.  here you go. output saved
> into attached file.

Thanks a lot!

> octave:5> pkg install -verbose parallel-3.1.3.tar.gz
> ...
> checking for gcc... gcc
> ...
> cc -c gnulib-wrappers.c -Igl -g -O2 -fPIC
> make: cc: Command not found

So configure finds gcc, but make uses 'cc'. I probably shouldn't make
CC configurable from environment in src/Makefile.in.

After fixing this (changeset 723cda56ca37), a new package with
pre-built docs has been uploaded, as before to:
https://sourceforge.net/p/octave/package-releases/393/
md5sum:
1fedf1ea96f8e249cd48421ef6a54a88  parallel-3.1.3.tar.gz

Olaf

--
public key id EAFE0591, e.g. on x-hkp://pool.sks-keyservers.net

sorry to be fumbling along here, where should I be able to find __test_rfeval__.cc? I see it in the pagkage tar.gz file, but am not finding it searching under the package folders. extracted it and tried to run it from another folder and got:
octave:9> test("__test_rfeval__.cc")
***** test
 __test_rfeval__ (@ (x) sqrt (x), 1, @ (info, x) info.message);
 assert (__test_rfeval__ (9, 1, @ (info, x) info.message){1}, 3);
 assert (iscellstr (__test_rfeval__ ("a", 1, @ (info, x) info.message)));
!!!!! test failed
'__test_rfeval__' undefined near line 3 column 2
octave:10>

 
Reply | Threaded
Open this post in threaded view
|

Re: [octave-forge] (parallel) request help for testing at Windows

PhilipNienhuis
In reply to this post by Olaf Till-2
Olaf Till-2 wrote

> Hi,
>
> while re-structuring the parallel package to not work with forked
> children of Octave sessions, I tried to make the local and client part
> of the package compatible with Windows. (The server part will not be
> built at windows.)
>
> Currently the infrastructure for the new code is in place and
> everything which is critical for Windows compatibility should be
> done. Preliminary tests pass at current Octave versions at
> GNU/Linux. The package cross-compiled successfully for mingw at
> mxe-octave.
>
> Could somebody run the tests described below at Windows? (Tests at
> other non-GNU OSs are welcome, too, of course.) Since building at
> Windows may need a package with pre-built docs, I uploaded one at:

I tried to "pkg install" but that didn't go very well.

Build log attached (octave-6.0.0. tip from Nov. 21)

Philip parallel-1.zip
<https://octave.1599824.n4.nabble.com/file/t248596/parallel-1.zip>  



--
Sent from: https://octave.1599824.n4.nabble.com/Octave-Maintainers-f1638794.html

Reply | Threaded
Open this post in threaded view
|

Re: [octave-forge] (parallel) request help for testing at Windows

Olaf Till-2
In reply to this post by nrjank
Thanks for going on with this...

On Sat, Nov 23, 2019 at 05:09:26PM -0500, Nicholas Jankowski wrote:
> where should I be able to find
> __test_rfeval__.cc? I see it in the pagkage tar.gz file, but am not finding
> it searching under the package folders.

I think the package tar.gz file is indeed the only place where
__test_rfeval__.cc is kept.

> extracted it and tried to run it
> from another folder and got:
> octave:9> test("__test_rfeval__.cc")
> ***** test
>  __test_rfeval__ (@ (x) sqrt (x), 1, @ (info, x) info.message);
>  assert (__test_rfeval__ (9, 1, @ (info, x) info.message){1}, 3);
>  assert (iscellstr (__test_rfeval__ ("a", 1, @ (info, x) info.message)));
> !!!!! test failed
> '__test_rfeval__' undefined near line 3 column 2
> octave:10>
Assuming that installation of the package now succeeded (did it?),
__test_rfeval__.oct, where the function '__test_rfeval__' is defined,
should be installed, too. The test didn't find it in the search path,
though -- maybe you didn't load the package with 'pkg load parallel'?

Olaf

--
public key id EAFE0591, e.g. on x-hkp://pool.sks-keyservers.net

signature.asc (849 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: [octave-forge] (parallel) request help for testing at Windows

Olaf Till-2
In reply to this post by PhilipNienhuis
On Sat, Nov 23, 2019 at 04:51:41PM -0600, PhilipNienhuis wrote:

> Olaf Till-2 wrote
> > Hi,
> >
> > while re-structuring the parallel package to not work with forked
> > children of Octave sessions, I tried to make the local and client part
> > of the package compatible with Windows. (The server part will not be
> > built at windows.)
> >
> > Currently the infrastructure for the new code is in place and
> > everything which is critical for Windows compatibility should be
> > done. Preliminary tests pass at current Octave versions at
> > GNU/Linux. The package cross-compiled successfully for mingw at
> > mxe-octave.
> >
> > Could somebody run the tests described below at Windows? (Tests at
> > other non-GNU OSs are welcome, too, of course.) Since building at
> > Windows may need a package with pre-built docs, I uploaded one at:
>
> I tried to "pkg install"
Thanks.

You know, I'm not very familiar with package installation at Windows
and must leave the choice of installation method to you...

> but that didn't go very well.

> Build log attached (octave-6.0.0. tip from Nov. 21)
>
> Philip parallel-1.zip
> <https://octave.1599824.n4.nabble.com/file/t248596/parallel-1.zip>  

Sorry, my fault --- I should have mentioned that the recent Octave
versions against which I tested did not include the development
branch, so failing building against Octaves development branch is
expected...

Olaf

--
public key id EAFE0591, e.g. on x-hkp://pool.sks-keyservers.net

signature.asc (849 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: [octave-forge] (parallel) request help for testing at Windows

nrjank
In reply to this post by Olaf Till-2
 maybe you didn't load the package with 'pkg load parallel'?


ummm, heh, yeah. that was it.  

so trying again:

octave:7> pkg load parallel
octave:8> test("__test_rfeval__.cc","verbose")
>>>>> c:\programs\octave\__test_rfeval__.cc
***** test
 __test_rfeval__ (@ (x) sqrt (x), 1, @ (info, x) info.message);
 assert (__test_rfeval__ (9, 1, @ (info, x) info.message){1}, 3);
 assert (iscellstr (__test_rfeval__ ("a", 1, @ (info, x) info.message)));
open: No such file or directory
process generation failed


so far octave has been sitting at this state for a couple hours, windows indicates all octave sub-processes at 0%


Reply | Threaded
Open this post in threaded view
|

Re: [octave-forge] (parallel) request help for testing at Windows

Olaf Till-2
On Sun, Nov 24, 2019 at 11:55:08AM -0500, Nicholas Jankowski wrote:

> octave:7> pkg load parallel
> octave:8> test("__test_rfeval__.cc","verbose")
> >>>>> c:\programs\octave\__test_rfeval__.cc
> ***** test
>  __test_rfeval__ (@ (x) sqrt (x), 1, @ (info, x) info.message);
>  assert (__test_rfeval__ (9, 1, @ (info, x) info.message){1}, 3);
>  assert (iscellstr (__test_rfeval__ ("a", 1, @ (info, x) info.message)));
> open: No such file or directory
> process generation failed
>
>
> so far octave has been sitting at this state for a couple hours, windows
> indicates all octave sub-processes at 0%
Do I understand it correctly that Octave didn't return to the prompt
after this? Even after hitting ENTER?

Anyway, the problem seems to be that the default name of the Octave
executable file or the path to this file is wrong. It occurs to me now
that I didn't consider the '.exe' postfix of Windows... But for the
case that it is not so simple, could you give me the following
information, still:

- From the Octave prompt, what das 'getenv("OCTAVE_BINDIR")' print out
  (if anything)?

- From the shell, what does  'octave-config -p BINDIR' say? (Maybe you
  must give the path to octave-config.)

- In the directories, if any, printed out by the above, what are the
  filenames of the Octave executables? Is there an 'octave-cli.exe'?

Thanks,

Olaf

--
public key id EAFE0591, e.g. on x-hkp://pool.sks-keyservers.net

signature.asc (849 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: [octave-forge] (parallel) request help for testing at Windows

nrjank
Do I understand it correctly that Octave didn't return to the prompt
after this? Even after hitting ENTER?

correct. had to kill octave from the task manager.
 
Anyway, the problem seems to be that the default name of the Octave
executable file or the path to this file is wrong. It occurs to me now
that I didn't consider the '.exe' postfix of Windows... But for the
case that it is not so simple, could you give me the following
information, still:

- From the Octave prompt, what das 'getenv("OCTAVE_BINDIR")' print out
  (if anything)?

 octave:1> getenv("OCTAVE_BINDIR")
ans =
octave:2>


- From the shell, what does  'octave-config -p BINDIR' say? (Maybe you
  must give the path to octave-config.)

not surprisingly 
"'octave-config' is not recognized as an internal or external command,
operable program or batch file."

a search below the Octave main folder for octave-config* only found:

C:\Programs\Octave\octave-5.1.0-w64\mingw64\share\man\man1\octave-config.1
C:\Programs\Octave\octave-5.1.0-w64\mingw64\include\octave-5.1.0\octave\octave-config.h


- In the directories, if any, printed out by the above, what are the
  filenames of the Octave executables? Is there an 'octave-cli.exe'?

obviously no output from above, but octave-cli* finds:
C:\Programs\Octave\octave-5.1.0-w64\mingw64\bin\octave-cli.exe
C:\Programs\Octave\octave-5.1.0-w64\mingw64\bin\octave-cli-5.1.0.exe


Reply | Threaded
Open this post in threaded view
|

Re: [octave-forge] (parallel) request help for testing at Windows

mmuetzel
In reply to this post by Olaf Till-2
Am 25. November 2019 um 11:15 Uhr schrieb "Olaf Till":

> On Sun, Nov 24, 2019 at 11:55:08AM -0500, Nicholas Jankowski wrote:
> > octave:7> pkg load parallel
> > octave:8> test("__test_rfeval__.cc","verbose")
> > >>>>> c:\programs\octave\__test_rfeval__.cc
> > ***** test
> >  __test_rfeval__ (@ (x) sqrt (x), 1, @ (info, x) info.message);
> >  assert (__test_rfeval__ (9, 1, @ (info, x) info.message){1}, 3);
> >  assert (iscellstr (__test_rfeval__ ("a", 1, @ (info, x) info.message)));
> > open: No such file or directory
> > process generation failed
> >
> >
> > so far octave has been sitting at this state for a couple hours, windows
> > indicates all octave sub-processes at 0%
>
> Do I understand it correctly that Octave didn't return to the prompt
> after this? Even after hitting ENTER?
>
> Anyway, the problem seems to be that the default name of the Octave
> executable file or the path to this file is wrong. It occurs to me now
> that I didn't consider the '.exe' postfix of Windows... But for the
> case that it is not so simple, could you give me the following
> information, still:
>
> - From the Octave prompt, what does 'getenv("OCTAVE_BINDIR")' print out
>   (if anything)?

The batch files set up the environment variable "OCTAVE_HOME". That points to a directory that looks like the Linux root directory. That is, the binaries are located in 'fullfile(getenv("OCTAVE_HOME"), "bin")'.
In my case OCTAVE_HOME is "C:\Octave\Octave-5.1.0.0\mingw64".

> - From the shell, what does  'octave-config -p BINDIR' say? (Maybe you
>   must give the path to octave-config.)

PS C:\Octave\Octave-5.1.0.0\mingw64\bin> .\octave-config.exe -p BINDIR
C:\Octave\Octave-5.1.0.0\mingw64\bin

> - In the directories, if any, printed out by the above, what are the
>   filenames of the Octave executables? Is there an 'octave-cli.exe'?

Yes, there is. But it is recommended to call the batch files instead. Just calling "octave" should be correct unless you want to specifically omit setting up the correct environment.
Once Octave is running, the "bin" directory is in the path:
>> system("octave-config -p BINDIR")
C:\Octave\OCTAVE~1.0\mingw64\bin
ans = 0
>>

HTH,
Markus

Reply | Threaded
Open this post in threaded view
|

Re: [octave-forge] (parallel) request help for testing at Windows

nrjank
> - From the shell, what does  'octave-config -p BINDIR' say? (Maybe you
>   must give the path to octave-config.)

PS C:\Octave\Octave-5.1.0.0\mingw64\bin> .\octave-config.exe -p BINDIR
C:\Octave\Octave-5.1.0.0\mingw64\bin



well, sure enough I do have octave-config hiding in there. once again, bit by the inane MS Windows search capability. 

search for octave-config*:  turns up octave-config.1 and octave-config.h
search for  octave-config.*:  now also includes octave-config.exe

*head-desk*

in any case, matching Markus's output:

C:\Programs\Octave\octave-5.1.0-w64\mingw64\bin>octave-config -p BINDIR
C:\Programs\Octave\octave-5.1.0-w64\mingw64\bin
Reply | Threaded
Open this post in threaded view
|

Re: [octave-forge] (parallel) request help for testing at Windows

Olaf Till-2
On Mon, Nov 25, 2019 at 05:28:13PM +0100, "Markus Mützel" wrote:

> Am 25. November 2019 um 11:15 Uhr schrieb "Olaf Till":
> ...
> > - From the shell, what does  'octave-config -p BINDIR' say? (Maybe you
> >   must give the path to octave-config.)
>
> PS C:\Octave\Octave-5.1.0.0\mingw64\bin> .\octave-config.exe -p BINDIR
> C:\Octave\Octave-5.1.0.0\mingw64\bin
>
> > - In the directories, if any, printed out by the above, what are the
> >   filenames of the Octave executables? Is there an 'octave-cli.exe'?
>
> Yes, there is. But it is recommended to call the batch files
> instead. Just calling "octave" should be correct unless you want to
> specifically omit setting up the correct environment.
> ...
On Mon, Nov 25, 2019 at 12:32:41PM -0500, Nicholas Jankowski wrote:
> in any case, matching Markus's output:
>
> C:\Programs\Octave\octave-5.1.0-w64\mingw64\bin>octave-config -p BINDIR
> C:\Programs\Octave\octave-5.1.0-w64\mingw64\bin

Thank you Markus and Nicholas,

I couldn't go on with this for a few days, and since it seems to
require more effort, I suggest that you don't fell pressed to respond
immediately but may be make a new test once you have time... I can't
currently promise to immediately respond myself...

Anyway a new tarball with md5sum

6624809226b50b3f2e475e98d84bad40  parallel-3.1.3.tar.gz

is uploaded to

https://sourceforge.net/p/octave/package-releases/393/

, which addresses the issue with the filename of the Octave
executable.

As for starting from batch files: The new process is started by
package 'parallel' with 'CreateProcessA()' in a way that sets up the
environment of the new process with the environment of the parent
process (Octave). So I don't think it's necessary to start the new
process indirectly with a batch file. The rationale in starting
octave-cli.exe with a fully specified path is that this new process
should be of the same Octave distribution as the parent process
(assuming it's possible to have several different Octave versions
installed, as in GNU/Linux).

Olaf

--
public key id EAFE0591, e.g. on x-hkp://pool.sks-keyservers.net

signature.asc (849 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: [octave-forge] (parallel) request help for testing at Windows

mmuetzel
Am 28. November 2019 um 19:20 Uhr schrieb "Olaf Till":

> I couldn't go on with this for a few days, and since it seems to
> require more effort, I suggest that you don't fell pressed to respond
> immediately but may be make a new test once you have time... I can't
> currently promise to immediately respond myself...
>
> Anyway a new tarball with md5sum
>
> 6624809226b50b3f2e475e98d84bad40  parallel-3.1.3.tar.gz
>
> is uploaded to
>
> https://sourceforge.net/p/octave/package-releases/393/
>
> , which addresses the issue with the filename of the Octave
> executable.

I downloaded the tar ball and installed it with
pkg install parallel-3.1.3.tar.gz

It looks like the compilation completed successfully.

Afterwards, I re-started Octave (don't know whether that is necessary) and tried to run the following command. But it looks like they failed:
octave:1> pkg load parallel
octave:2>  __test_rfeval__ (@ (x) sqrt (x), 1, @ (info, x) info.message);
error: '__rfeval_slave__' undefined near line 1 column 1
could not load variable
error: could not initialize function handle

The function seems to exist though:
octave:2> which __rfeval_slave__
'__rfeval_slave__' is a function from the file D:\Octave\Octave-5.1.0.0\mingw64\lib\octave\packages\parallel-3.1.3\x86_64-w64-mingw32-api-v53\__rfeval_slave__.oct

And on the powershell:
PS D:\Octave\Octave-5.1.0.0\mingw64\lib\octave\packages\parallel-3.1.3\x86_64-w64-mingw32-api-v53> ls


    Verzeichnis: D:\Octave\Octave-5.1.0.0\mingw64\lib\octave\packages\parallel-3.1.3\x86_64-w64-mingw32-api-v53


Mode                LastWriteTime         Length Name
----                -------------         ------ ----
-a----       30.11.2019     10:06        8298196 fload.oct
-a----       30.11.2019     10:06        8298713 fsave.oct
-a----       30.11.2019     10:06       16877329 parallel_generate_srp_data.oct
-a----       30.11.2019     10:06       39652934 parallel_interface.oct
-a----       30.11.2019     10:06            483 PKG_ADD
-a----       30.11.2019     10:06            573 PKG_DEL
-a----       30.11.2019     10:06        8448098 select.oct
-a----       30.11.2019     10:06        6728715 __exit__.oct
-a----       30.11.2019     10:06       11030765 __rfeval_slave__.oct
-a----       30.11.2019     10:06       15966986 __test_rfeval__.oct
-a----       30.11.2019     10:06        8669054 __visglobal__.oct


Please, let me know if you need more information.

Markus

Reply | Threaded
Open this post in threaded view
|

Re: [octave-forge] (parallel) request help for testing at Windows

Olaf Till-2
On Sat, Nov 30, 2019 at 10:26:51AM +0100, "Markus Mützel" wrote:

> Am 28. November 2019 um 19:20 Uhr schrieb "Olaf Till":
> > I couldn't go on with this for a few days, and since it seems to
> > require more effort, I suggest that you don't fell pressed to respond
> > immediately but may be make a new test once you have time... I can't
> > currently promise to immediately respond myself...
> >
> > Anyway a new tarball with md5sum
> >
> > 6624809226b50b3f2e475e98d84bad40  parallel-3.1.3.tar.gz
> >
> > is uploaded to
> >
> > https://sourceforge.net/p/octave/package-releases/393/
> >
> > , which addresses the issue with the filename of the Octave
> > executable.
>
> I downloaded the tar ball and installed it with
> pkg install parallel-3.1.3.tar.gz
>
> It looks like the compilation completed successfully.
>
> Afterwards, I re-started Octave (don't know whether that is necessary) and tried to run the following command. But it looks like they failed:
> octave:1> pkg load parallel
> octave:2>  __test_rfeval__ (@ (x) sqrt (x), 1, @ (info, x) info.message);
> error: '__rfeval_slave__' undefined near line 1 column 1
At least we are one step further, the child process at least started
(and then exited with an error).

I may have to let 'pkg load ...' called on the commandline of the
child process, but for now, could you insert a:

pkg load parallel

into your '.octaverc' file and re-run the test? The file may be named
slightly differently on Windows...

> could not load variable
> error: could not initialize function handle
>

These errors won't probably disappear by this, though...

Olaf

--
public key id EAFE0591, e.g. on x-hkp://pool.sks-keyservers.net

signature.asc (849 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: [octave-forge] (parallel) request help for testing at Windows

mmuetzel
Am 30. November 2019 um 21:19 Uhr schrieb "Olaf Till"

> At least we are one step further, the child process at least started
> (and then exited with an error).
>
> I may have to let 'pkg load ...' called on the commandline of the
> child process, but for now, could you insert a:
>
> pkg load parallel
>
> into your '.octaverc' file and re-run the test? The file may be named
> slightly differently on Windows...
>
> > could not load variable
> > error: could not initialize function handle
> >
>
> These errors won't probably disappear by this, though...

That seems to have done the trick:

>> test("__test_rfeval__.cc", "verbose")
>>>>> D:\Repositories\Octave\of-parallel\src\__test_rfeval__.cc
***** test
 __test_rfeval__ (@ (x) sqrt (x), 1, @ (info, x) info.message);
 assert (__test_rfeval__ (9, 1, @ (info, x) info.message){1}, 3);
 assert (iscellstr (__test_rfeval__ ("a", 1, @ (info, x) info.message)));
error: sqrt: argument must be numeric
***** test
 __test_rfeval__ (@ (x) sqrt (x), 1, 1);
 assert (__test_rfeval__ (9, 1, 1){1}, 3);
 assert (__test_rfeval__ ("a", 1, 1), int32 (-1));
error: sqrt: argument must be numeric
PASSES 2 out of 2 tests
>>

Markus


Reply | Threaded
Open this post in threaded view
|

Re: [octave-forge] (parallel) request help for testing at Windows

Olaf Till-2
On Sat, Nov 30, 2019 at 11:02:46PM +0100, "Markus Mützel" wrote:

> Am 30. November 2019 um 21:19 Uhr schrieb "Olaf Till"
> > At least we are one step further, the child process at least started
> > (and then exited with an error).
> >
> > I may have to let 'pkg load ...' called on the commandline of the
> > child process, but for now, could you insert a:
> >
> > pkg load parallel
> >
> > into your '.octaverc' file and re-run the test? The file may be named
> > slightly differently on Windows...
> >
> > > could not load variable
> > > error: could not initialize function handle
> > >
> >
> > These errors won't probably disappear by this, though...
>
> That seems to have done the trick:
>
> >> test("__test_rfeval__.cc", "verbose")
> >>>>> D:\Repositories\Octave\of-parallel\src\__test_rfeval__.cc
> ***** test
>  __test_rfeval__ (@ (x) sqrt (x), 1, @ (info, x) info.message);
>  assert (__test_rfeval__ (9, 1, @ (info, x) info.message){1}, 3);
>  assert (iscellstr (__test_rfeval__ ("a", 1, @ (info, x) info.message)));
> error: sqrt: argument must be numeric
> ***** test
>  __test_rfeval__ (@ (x) sqrt (x), 1, 1);
>  assert (__test_rfeval__ (9, 1, 1){1}, 3);
>  assert (__test_rfeval__ ("a", 1, 1), int32 (-1));
> error: sqrt: argument must be numeric
> PASSES 2 out of 2 tests
> >>
>
> Markus
I'm delighted! Thanks very much for testing. So process creation and
process communication seem to work. I've not yet requested a test for
process termination, but this can probably be postponed till the rest
is ready, so that should be all for now.

Thanks again,

Olaf

--
public key id EAFE0591, e.g. on x-hkp://pool.sks-keyservers.net

signature.asc (849 bytes) Download Attachment