[Pkg-octave-devel] Bug#493869: Bug#493869: octave3.0: gaminv and gamcdf crash for big values of parameter A

[Sébastien Villemot]

"Thomas Weber" <thomas.weber.mail at gmail.com> a écrit :

> Am Dienstag, den 05.08.2008, 15:49 +0200 schrieb Sebastien Villemot:
>> Package: octave3.0
>> Version: 1:3.0.1-6
>> Severity: normal
>>
>> gaminv and gamcdf function crash with the following parameters:
>
>> octave> gaminv(0.5,600,1)
>>  ***MESSAGE FROM ROUTINE D9LGIT IN LIBRARY SLATEC.
>>  ***FATAL ERROR, PROG ABORTED, TRACEBACK REQUESTED
>>  *  NO CONVERGENCE IN 200 TERMS OF CONTINUED FRACTION
>>  *  ERROR NUMBER = 3
>>  *
>>  ***END OF MESSAGE
>
>> octave> gamcdf(600, 600, 1)
>>  ***MESSAGE FROM ROUTINE D9LGIT IN LIBRARY SLATEC.
>>  ***FATAL ERROR, PROG ABORTED, TRACEBACK REQUESTED
>>  *  NO CONVERGENCE IN 200 TERMS OF CONTINUED FRACTION
>>  *  ERROR NUMBER = 3
>>  *
>>  ***END OF MESSAGE
>
> That's not a crash, is it? The functions inform you that they didn't
> converge.

I called it a crash since the functions aborted without returning a  
meaningful value, but it doesn't crash Octave. Calling it a failure  
may therefore be more appropriate.

>> Note that if you replace 600 by 500, the problem doesn't occur.
>
> See
> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=418158
>
> d9lgit.f needs an improvement, either directly or by completely
> replacing it with something better.

Thanks for pointing me to this bug report. I hadn't noticed that the  
bug in poisscdf() was indeed caused by the same problem in gammainc(),  
itself caused by a problem in d9lgit.

Best

Sébastien Villemot