SWIG 2.0.5 does not work with PyDOLFIN

[Johan Hake]

The recent release of SWIG does not work with DOLFIN. Some sort of type conversion bug is introduced. Can be reproduced by running the test suite:

  make run_quicktest

[Johan Hake (johan-hake)]

This is caused by change in SWIG. SWIG_From_unsigned_SS_int now returns a python long. I have asked the SWIG devel list if this change is intended or a regression bug.

[Garth Wells (garth-wells)]

Thread on swig list:

http://sourceforge.net/mailarchive/message.php?msg_id=29233828

[Johan Hake (johan-hake)]

Pointer to an actual SWIG bug also relevant for this bug:

https://sourceforge.net/tracker/?func=detail&atid=101645&aid=3525050&group_id=1645

I will try to add workarounds for both of these issues.

[Johannes Ring (johannr)]

Is it much work to fix for this in the 1.0.x branch? SWIG 2.0.5 is in Debian unstable and I would like to add the fix to the DOLFIN package.

[Johan Hake (johan-hake)]

On 05/10/2012 01:10 PM, Johannes Ring wrote:
> Is it much work to fix for this in the 1.0.x branch? SWIG 2.0.5 is in
> Debian unstable and I would like to add the fix to the DOLFIN package.

Urk... Probably not. Most of the changes was done in typemap files which
are all more or less mirrored in 1.0.0. I have no time looking at that
right now. SWIG has already stolen too much of my time lately.

Johan

[Johannes Ring (johannr)]

OK, I'll try backporting your fix.

[Johan Hake (johan-hake)]

The bug with shadownig typemaps in SWIG has now been fixed, and to SWIG's creadit, it now supports templated types of typedefs much better. This bug was a regression introduced while fixing this in SWIG.

The error introduced by SWIG_From(insigned int) now returning a Python long instead of a Pythin int, is now reported as a bug:

https://sourceforge.net/tracker/?func=detail&aid=3526487&group_id=1645&atid=101645

[Anders Logg (logg)]

On Mon, May 14, 2012 at 07:48:35AM -0000, Johan Hake wrote:
> The bug with shadownig typemaps in SWIG has now been fixed, and to
> SWIG's creadit, it now supports templated types of typedefs much better.
> This bug was a regression introduced while fixing this in SWIG.
>
> The error introduced by SWIG_From(insigned int) now returning a Python
> long instead of a Pythin int, is now reported as a bug:
>
> https://sourceforge.net/tracker/?func=detail&aid=3526487&group_id=1645&atid=101645
>
> ** Bug watch added: SourceForge.net Tracker #3526487
> http://sourceforge.net/support/tracker.php?aid=3526487

I didn't know it was possible to watch sf bugs from lp. Nice!

--
Anders

[Johan Hake (johan-hake)]

On 05/14/2012 11:56 AM, Anders Logg wrote:
> On Mon, May 14, 2012 at 07:48:35AM -0000, Johan Hake wrote:
>> The bug with shadownig typemaps in SWIG has now been fixed, and to
>> SWIG's creadit, it now supports templated types of typedefs much better.
>> This bug was a regression introduced while fixing this in SWIG.
>>
>> The error introduced by SWIG_From(insigned int) now returning a Python
>> long instead of a Pythin int, is now reported as a bug:
>>
>> https://sourceforge.net/tracker/?func=detail&aid=3526487&group_id=1645&atid=101645
>>
>> ** Bug watch added: SourceForge.net Tracker #3526487
>> http://sourceforge.net/support/tracker.php?aid=3526487
>
> I didn't know it was possible to watch sf bugs from lp. Nice!

Nor, did I. Did someone add these bug trackers manually or did launchpad
automagically extract the bugs from the links?

Johan