Re: [Plplot-devel] [Perldl] My initial pre-release testing

[Alan W. I. <ir...@be...>]

On 2008-12-05 15:02-0700 Doug Hunt wrote:

> Hi Alan:
>
>> Could you double check with a fresh checkout and clean build of PLplot and 
>> a
>> fresh download and build of PDL-Graphics-PLplot-0.42.tar.gz?
>
> I did a fresh check-out and build of plplot and a fresh install of 
> PDL-Graphics-PLplot-0.42 and I still get the same (correct) output from
> x02.pl, which is attached.  Everything looks good on my end--the perl output 
> matches the c output.

It matches my C output here as well.

So it appears we have some sort of platform or PDL Graphics PLplot module
difference which is, of course, a pain in the ass to figure out.

Will someone else try testing Doug's updates, please, to see how widespread
the problem is?

>
> I've also uploaded a new version of the PDL interface 
> (PDL-Graphics-PLplot-0.43) to CPAN which allows one to specify the plplot 
> library and include areas as environment variables instead of having to hack 
> on the Makefile.PL, as per your question 1) from another email.
>
>> 
>> In case our different results are due to some platform difference I am
>> running Debian lenny with perl 5.10.0-17 and pdl 1:2.4.3-8 and following 
>> the
>> 0.42 update directions in examples/perl/README.perldemos.  Could you please
>> also review those update directions to be sure they are correct?
>
> I tried out the instructions you give in examples/perl/README.perldemos--the 
> 'make install' step does not work on my machine, so I came up with another 
> approach:  specify the install directory on the 'perl Makefile.PL' step 
> instead of the 'make install' step.  I think this is more correct perl-wise.
>
> I've updated the README.perldemos file to include specifying plplot location 
> in environment variables and specifying the install area on the 'perl 
> Makefile.PL' step.  I've not updated the SVN, but this file is attached.
>
> My guess is that you might not be running with the correct version of 
> PDL-Graphics-PLplot.  Could you try using the new instructions in 
> README.perldemos and seeing if things make more sense after that?

Sure, I will try.  However, that build only produces a relatively small
number of files in the install tree so there is not too much that can go
wrong with symlinking the files from that special install area to the
correct place in the system PDL location.  And it appears that method worked
fine to get example 20 in perfect shape.  So I am looking for some other
reason (see below) to explain the different results we get for example 2,
but not for example 20 (which you fixed later).

Are your present results for a 32-bit system or a 64-bit system?  Mine is
64-bit.  If you have only tested on a 32-bit system, could you (and other
PDL testers here as well, I hope) test on a 64-bit system as well?  The
agreement we get for the C examples shows there are no problems with example
2 for the core C library.  Thus, the disagreement for the PDL example 2
could be explained either by a 64-bit bug in the second standard example
implemented in PDL or by a 64-bit bug in PDL::Graphics::PLplot that is
exposed by that example.

I will also look at example 8 (probably tomorrow).

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state implementation
for stellar interiors (freeeos.sf.net); PLplot scientific plotting software
package (plplot.org); the libLASi project (unifont.org/lasi); the Loads of
Linux Links project (loll.sf.net); and the Linux Brochure Project
(lbproject.sf.net).
__________________________

Linux-powered Science
__________________________