Re: [Plplot-devel] Octave 4.0.0

[Orion P. <or...@co...>]

On 07/08/2015 03:36 PM, Alan W. Irwin wrote:
> On 2015-07-08 15:46-0000 Arjen Markus wrote:
> 
>> Well, I tried it with Octave allowed, but I get build errors - see
> the tarball. It does not seem to be complaining about 2 versus 4
> arguments anymore, but something is definitely wrong. Could that be
> the SWIG version? (Note: I did not use the very latest checkout, but
> the commits I saw have nothing to do with Octave).
> 
> To Arjen and Orion:
> 
> @Arjen:
> Thanks for this comprehensive test report.
> 
> The short response for working around the octave error you
> encountered is you should remove the Cygwin package
> octave-devel-4.0.0-1 and install instead octave-devel-3.8.2-1. (And
> you should also confirm you have libhdf5-devel-1.8.15-1 installed.)
> And then you should try the test again (probably it is best if you do
> that with the PLplot master tip version).  I am pretty sure that
> will "just work" since the 2 versus 4 argument issue has now been fixed
> and we have plenty of good tests of octave-3 on other platforms.
> 
> @Orion:
> 
> I suggest you also use Octave-3 for now assuming that Fedora has
> Octave-3 packages.
> 
> @Arjen:
> 
> Here is my longer response to your report.
> 
> It turns out you ran into virtually exactly the same octave problems
> that Orion reported here so you may want to reread his recent post
> about this and my response to it.
> 
> The first part of Orion's post shows that the swig-3.0.6
> octave component does not like leading blanks in documentation of the
> swig-generated bindings for some reason, but swig-3.0.5 is OK with
> that. I have fixed that issue as of the current master tip.  Your use
> of ae0e9da misses that fix to our documentation of the swig-generated
> bindings, but missing that fix did not matter for your case since
> Cygwin currently supplies you with swig-3.0.5.
> 
> However, once Orion locally fixed the leading blank issue in our
> documentation of swig-generated bindings (the issue that is now
> permanently fixed in master tip), he went on to describe additional
> issues with octave #includes supplied by swig that are not compatible
> with Octave-4.0.0, and it appears you ran into those exact same
> swig issues.
> 
> In sum, it appears build issues like you encountered above will not allow us
> to even try Octave-4.0.0 until swig fixes their Octave-4.0.0
> #include issues, and that swig fix propagates to swig versions we are
> using.  So for now (commit id 2ad5aef) the build system issues a
> WARNING if it finds Octave-4.  For that case it disables the octave
> bindings unless the user specifically opts in using the cmake
> experimental option -DTRY_OCTAVE4=ON.
> 
> Once swig fixes their problem with Octave-4 #includes so that
> -DTRY_OCTAVE4=ON has even a chance of working, I assume we will have a
> lot of work ahead of us both in the octave bindings and octave
> examples to become compatible with Octave-4.  So I expect
> -DTRY_OCTAVE4=ON will be experimental for quite some time after the
> required swig fix.
> 
> Alan
> __________________________
> Alan W. Irwin

My fix for octave 4.0,0 support in swig is here:

https://github.com/swig/swig/pull/460

I'll test builds of plplot once the swig build is complete.


-- 
Orion Poplawski
Technical Manager                     303-415-9701 x222
NWRA, Boulder/CoRA Office             FAX: 303-415-9702
3380 Mitchell Lane                       or...@nw...
Boulder, CO 80301                   http://www.nwra.com