Re: [Quantlib-dev] [Quantlib-users] Preliminary tarballs for QuantLib 1.7

[Peter C. <pca...@gm...>]

Hi Klaus,

it's better with your fix, but still not perfect:

==22028== Invalid read of size 8
==22028==    at 0x559A5AD: __normal_iterator (stl_iterator.h:729)
==22028==    by 0x559A5AD: end (stl_vector.h:566)
==22028==    by 0x559A5AD:
QuantLib::FdmExpExtOUInnerValueCalculator::innerValue(QuantLib::FdmLinearOpIterator
const&, double) (fdmexpextouinnervaluecalculator.hpp:55)
==22028==    by 0x59CD89C:
QuantLib::FdmSimpleStorageCondition::applyTo(QuantLib::Array&, double)
const (fdmsimplestoragecondition.cpp:77)
==22028==    by 0x59D175E:
QuantLib::FdmStepConditionComposite::applyTo(QuantLib::Array&, double)
const (fdmstepconditioncomposite.cpp:63)
==22028==    by 0x59D175E:
QuantLib::FdmStepConditionComposite::applyTo(QuantLib::Array&, double)
const (fdmstepconditioncomposite.cpp:63)
==22028==    by 0x599416D: rollbackImpl (finitedifferencemodel.hpp:103)
==22028==    by 0x599416D: rollback (finitedifferencemodel.hpp:86)
==22028==    by 0x599416D:
QuantLib::FdmBackwardSolver::rollback(QuantLib::Array&, double,
double, unsigned long, unsigned long) (fdmbackwardsolver.cpp:117)
==22028==    by 0x597DD48:
QuantLib::Fdm2DimSolver::performCalculations() const
(fdm2dimsolver.cpp:78)
==22028==    by 0x597B186: calculate (lazyobject.hpp:150)
==22028==    by 0x597B186:
QuantLib::Fdm2DimSolver::interpolateAt(double, double) const
(fdm2dimsolver.cpp:88)
==22028==    by 0x55DD743: valueAt (fdmsimple2dextousolver.hpp:53)
==22028==    by 0x55DD743:
QuantLib::FdSimpleExtOUStorageEngine::calculate() const
(fdsimpleextoustorageengine.cpp:181)
==22028==    by 0x4E3FCE: QuantLib::Instrument::performCalculations()
const (instrument.hpp:168)
==22028==    by 0x4F684E: calculate (lazyobject.hpp:150)
==22028==    by 0x4F684E: calculate (instrument.hpp:153)
==22028==    by 0x4F684E: QuantLib::Instrument::NPV() const (instrument.hpp:187)
==22028==    by 0xD69C79: VPPTest::testSimpleExtOUStorageEngine() (vpp.cpp:254)

It seems that t is not initialized in fdmexpextouinnervaluecalculator.hpp:55.

Best regards
Peter

On 15 November 2015 at 23:39, Klaus Spanderen <kl...@sp...> wrote:
> Hi Peter,
>
> thanks for highlighting this issue. You're right, the test compares
> uninitialized values with themselves and therefore the explanatory power is
> not as good as it could be. IMO the fix is a one line change and I've sent a PR
> to Luigi.
>
> regards
> Klaus
>
>
> On Saturday, November 14, 2015 06:04:04 PM Peter Caspers wrote:
>> I tested QuantLib-1.7.tar.gz, it builds fine with gcc 4.9.3 / boost
>> 1.59.0 on my machine. The test-suite runs fine as well, also under
>> memcheck, except for a use of uninitialized values triggered from
>>
>> VPPTest::testKlugeExtOUMatrixDecomposition()
>>
>> but I do not think that this is specific to the release. Maybe Klaus,
>> can you have a closer look ? I attach the full memcheck log.
>>
>> Two more observations (also unrelated to 1.7, anyway):
>>
>> The following files are wrongly marked as executable, is that coming
>> from the repository?
>>
>> -rwxrwxr-x 1 peter peter   3085 Nov 12 15:38 ql/userconfig.hpp
>> -rwxrwxr-x 1 peter peter   7844 Nov 12 15:38 test-suite/observable.cpp
>> -rwxrwxr-x 1 peter peter  14561 Nov 12 15:38
>> test-suite/quantlibtestsuite.cpp
>>
>> Second thing is that when running the test-suite overnight and the
>> system date changes in the middle of a test, the check for restoring
>> the settings (if applicable to the case) may fail, like here
>>
>> utilities.hpp(80): error: in "QuantLib test suite/Forward option
>> tests/QuantLib::detail::quantlib_test_case(&ForwardOptionTest::testGreeksIni
>> tialization)": Evaluation date not reset
>>   before: November 13th, 2015
>>   after:  November 14th, 2015
>>
>> I guess that would be fixed though by the more general attempt to make
>> the suite deterministic w.r.t. the evaluation date (PR #313).
>>
>> Peter
>>
>> On 13 November 2015 at 17:11, Luigi Ballabio <lui...@gm...>
> wrote:
>> > Hi all,
>> >
>> >     new tentative tarballs (which by now should be pretty close to release
>> >
>> > 1.7, and include a list of changes) are available at
>> > <https://sourceforge.net/projects/quantlib/files/QuantLib/prerelease/>. if
>> > you have some time, please try them out and report any problems.
>> >
>> > Thanks,
>> >
>> >     Luigi
>> >
>> > On Fri, Oct 30, 2015 at 5:04 PM Luigi Ballabio <lui...@gm...>
>> >
>> > wrote:
>> >> Hi all,
>> >>
>> >>     tarballs of what will eventually become QuantLib 1.7 are available at
>> >>
>> >> <https://sourceforge.net/projects/quantlib/files/QuantLib/prerelease/>;
>> >> if
>> >> you have some time, please try them out and report any problems. The list
>> >> of changes is not yet available, but a list of the pull requests that
>> >> made it into this release are at
>> >> <https://github.com/lballabio/quantlib/issues?utf8=%E2%9C%93&q=milestone%
>> >> 3A%221.7+release%22+>
>> >>
>> >> Thanks,
>> >>
>> >>     Luigi
>> >>
>> >> --
>> >>
>> >> <http://leanpub.com/implementingquantlib>
>> >> <http://implementingquantlib.com>
>> >> <http://twitter.com/lballabio>
>> >
>> > --
>> >
>> > <http://leanpub.com/implementingquantlib>
>> > <http://implementingquantlib.com>
>> > <http://twitter.com/lballabio>
>> >
>> >
>> > --------------------------------------------------------------------------
>> > ----
>> >
>> > _______________________________________________
>> > QuantLib-dev mailing list
>> > Qua...@li...
>> > https://lists.sourceforge.net/lists/listinfo/quantlib-dev
>