Mon 16 May 2016 05:09:51 PM UTC, comment #11:
> I do not see other ways of "failing gracefully" ...
I don't know, perhaps can it be achieved through the Dependencies fields in DESCRIPTION?
Or perhaps using a preprocessor directive such as
?
|
Mon 16 May 2016 05:06:30 PM UTC, comment #10:
some things will definitely not work in 4.0.2, e.g. sending/receiving scalar structures and cell arrays.
what about bumping the Octave dependecy to 4.1 so that
it won't install in 4.0?
I do not see other ways of "failing gracefully" ...
|
Mon 16 May 2016 04:37:40 PM UTC, comment #9:
Well, I have no idea if it "works reasonably" in 4.0.2.
But currently it cannot even be installed, and installation fails with a rather cryptic error message (see my initial post)...
Is there any reason to believe that it doesn't work in 4.0.2?
Is there a test suite that I could run to know about that?
If it cannot be made to run in 4.0.x soon, perhaps should you consider a 1.2.1 release that simply fails more gracefully when a user tries to install it from 4.0.x?
|
Mon 16 May 2016 02:39:49 PM UTC, comment #8:
There's more issues that need fixing for making MPI
portable to Octave 4.2, e.g. bug #45294 .
I did not plan to support both 4.0.x and 4.2 versions,
but if you think the version currently in the repository
works reasonably in 4.0.2 and is worth releasing we could
make a release right now.
|
Mon 16 May 2016 08:40:31 AM UTC, comment #7:
The patch has been applied in august 2015 and the bug report has been closed...
But there has been no release of package mpi since then, so the problem still exists for people that install the package using "pkg install -forge mpi"
:-(
|
Wed 09 Sep 2015 10:22:28 PM UTC, comment #6:
@Carlo
It seems that the patch is not available any more. I found the same problem but I couldn't find the patch.
Best
Luiz
|
Mon 17 Aug 2015 05:23:34 PM UTC, comment #5:
I applied the patch with preprocessing directives as suggested.
thanks,
c.
|
Mon 08 Jun 2015 07:56:41 AM UTC, comment #4:
@Carlo: perhaps should you use preprocessing directives, as Mike Miller did for the communications package, for the sake of backward compatibility ?
http://sourceforge.net/p/octave/communications/ci/dd3e0dc1cd053909153fb03e5c79dc19f6068f4a/
|
Mon 08 Jun 2015 07:53:21 AM UTC, comment #3:
---- Posting here reply to Michael Creel on the mailing list
>> Carlo, that patch solved the problem.
OK, thanks for testing, so I'll apply the pacth and prepare a release.
>> I can't log into the bug tracker
>> (though I have not changed my password) so I have not added a
>> comment there.
Even though you have not changed your password you may want to try "[Lost your password?]"
on the login screen to recover your account info on savannah, although unlikely, it may be
that someone guessed your password and changed it ...
>> One other note, the mc_example.m in the inst directory of the
>> mpi package
>> does not run with Octave v4.0. If the olswrapper function is
>> pulled out and
>> put in a separate file, the example will run.
I'll look into this before the release but at first sight this sounds like an interpreter
bug rather than a problem with the mpi package ...
|
Fri 05 Jun 2015 04:14:53 PM UTC, comment #2:
I don't have a working installation of Octave 4.0
but the fix should be trivial, can you try the attached patch?
c.
(file #34166)
|
Fri 05 Jun 2015 02:21:22 PM UTC, comment #1:
adding maintainer of the mpi package
|
Fri 05 Jun 2015 11:35:06 AM UTC, original submission:
mpi-1.2.0 build fails in Octave 4.0.0
The problem is similar to
https://savannah.gnu.org/bugs/?func=detailitem&item_id=44063
Here is the log:
|