Re: Problem testing [Re: Final tarballs]

Posted by Liguo Song on Aug 28, 2003; 8:30am
URL: http://quantlib.414.s1.nabble.com/Final-tarballs-tp10320p10323.html

I finally found the trouble making optimization flag. It is
-ffast-math.

So, the test-suite will fail on the Testing Monte Carlo European
engines against analytic results with error:
 "McSimulation::value : max number of samples exceeded"

I am not sure whether it is due to the libraries or the test-suite
itself. I am going to leave out this troublesome optimization flag.

Hope someone else can learn from my mistake. :)

The final version RPM packages will be ready soon.


Liguo



On Wed, 27 Aug 2003, Liguo Song wrote:

> Hi, Dirk,
>
> I was just pulling my hair off for another problem with quantlib-test-suite. My
> RPM packages build fine, but when running quantlib-test-suite, it runs through
> all the tests, but ends up with the following error message:
>
>      !!!FAILURES!!!
>      Test Results:
>      Run:  64   Failures: 0   Errors: 1
>
>
>      1) test: Testing Monte Carlo European engines against analytic results (E)
>       "McSimulation::value : max number of samples exceeded"
>
>
> Right now, the quantlib-test-suite is a little out of my reach. Any input?
>
> BTW, if it helps, I am building the final tar balls that Luigi mentioned
> previously.
>
> Liguo
>
>
> Dirk Eddelbuettel wrote:
> > Using the exact same settings, compiler versions, ... as yesterday on the
> > previous golden master, I now fail when running make check.  I first noticed
> > it from the build log and, to exclude any terminal settings issues, re-ran it.
> >
> > It dies here:
> >
> > /usr/bin/make  check-TESTS
> > make[3]: Entering directory home/edd/src/debian/QuantLib-0.3.3/test-suite'
> >                    
> >                     Testing can never demonstrate the absence
> >                    of errors in software, only their presence.
> >                                            -- W.E. Dijkstra
> >  
> > ======================
> > Testing QuantLib 0.3.3
> > ======================
> > Testing joint calendars...
> > Testing cap/floor dependency on
> > strike...lt-quantlib-test-suite: relocation error: lt-quantlib-test-suite:
> > undefined symbol:
> > _ZN8QuantLib9CashFlows24FloatingRateCouponVectorERKSt6vectorIdSaIdEERKNS_4DateES8_iRKNS_8CalendarENS_17RollingConventionERKNS_6HandleINS_7Indexes5XiborEEEiS5_S8_
> > FAIL: quantlib-test-suite
> > ===================================================
> > 1 of 1 tests failed
> > Please report to [hidden email]
> > ===================================================
> >
> > I will hold back the upload til I here from Luigi. As I noted, all 62 of
> > these passed with flying colours yesterday.
> >
> > Liguo: I guess we could compare g++ et al version off-line if this works
> > for you. I am running Debian unstable, g++ is at 3.3.2-0pre1.
> >
> > Dirk
> >
>
>
>
>
> -------------------------------------------------------
> This sf.net email is sponsored by:ThinkGeek
> Welcome to geek heaven.
> http://thinkgeek.com/sf
> _______________________________________________
> Quantlib-dev mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/quantlib-dev
>