dolfin team mailing list archive

["Garth N. Wells" <gnw20@xxxxxxxxx>]

Niclas Jansson wrote:
> Anders Logg <logg@xxxxxxxxx> writes:
> 
>> On Wed, Feb 10, 2010 at 09:31:06AM +0000, Garth N. Wells wrote:
>>> I get occasionally get a CGAL error when running DOLFIN:
>>>
>>> terminate called after throwing an instance of 'CGAL::Assertion_exception'
>>>   what():  CGAL ERROR: assertion violation!
>>> Expr: -CGAL_IA_MUL(-1.1, 10.1) != CGAL_IA_MUL(1.1, 10.1)
>>> File: /home/garth/local/include/CGAL/Interval_nt.h
>>> Line: 158
>>> Explanation: Wrong rounding: did you forget the -frounding-math option
>>> if you use GCC?
>>> ==10917==
>>>
>>> It typically arises when trying to run inside a debugger, etc. Has
>>> anyone else seen this? I do compile with '-frounding-math'.
>>>
>>> Garth
>> Yes, I've seen this too when trying to debug some code. It's quite
>> annoying. It's somewhat similar to PETSc in that CGAL seems to go in
>> and catch errors and prevent debugging. I don't know what to do about
>> it.
>>
>> --
>> Anders
> 
> 
> You can disable these checks by recompiling CGAL,
> http://www.cgal.org/Manual/3.3/doc_html/Developers_manual/Developers_manual/Chapter_checks.html#Section_5.2
>

Thanks.

I don't think that I had the checks on since I made a 'release' build,
but CGAL is heavily templated so we probably need some options in the
DOLFIN build system to control CGAL checks in templated code (like we do
for uBLAS). The flags for the latest version can be found here:

http://www.cgal.org/Manual/3.5/doc_html/cgal_manual/Preliminaries/Chapter_main.html

(Section 2.10)

Garth

> As for PETSc, the command line argument '-no_signal_handler' removes
> most of the annoying stuff.
> 
> Niclas
> 
>