[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Bug-gsl] gsl 2.3.0 bug?

From: Rico Buchholz
Subject: [Bug-gsl] gsl 2.3.0 bug?
Date: Tue, 20 Nov 2018 17:36:58 +0100
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101 Thunderbird/60.3.0


I am working with a fortran project which uses fgsl (v1.2.0, together with gsl v2.3.0).

The code reports an arithmetic exception in an integration routine, when trying to integrate exp(-x^2) from x=19.3 to 20.0. I would expect the code to work without problems (maybe underflow or denormal). Trying to create a minimal working example i came up with the attached file (the cmake input i used for compilation is also attached, sorry I could not boil this further down). As a corresponding version for c++ works (also attached), this might relate to a bug in fgsl, but for me (a person without knowledge of the internals), that seems questionable according to what causes the problem and what not.

The output is as follows:

Program received signal SIGFPE: Floating-point exception - erroneous arithmetic operation.

Backtrace for this error:
#0  0x7fa39acc5d1d in ???
#1  0x7fa39acc4f7d in ???
#2  0x7fa39a1e405f in ???
#3  0x560ac54c9eff in ???
#4  0x560ac54860cf in neo2
    at /afs/itp.tugraz.at/user/buchholz/Programs/neo2-minimal/NEO-2-QL/neo2.f90:31
#5  0x560ac54861d8 in main
    at /afs/itp.tugraz.at/user/buchholz/Programs/neo2-minimal/NEO-2-QL/neo2.f90:4

It does not occur with the alternative integration routine commented in the code. This does not occur, if one reduces the lower bound to 19.2 or smaller (which is a reason why I suspect that this might be a gsl bug - a bug in the interface should have no influence on this).

With kind regards
Rico Buchholz

PS: I did also send the bug report to fgsl.

Attachment: CMakeLists.txt
Description: Text document

Attachment: neo2.f90
Description: Text Data

Attachment: main.cpp
Description: Text Data

reply via email to

[Prev in Thread] Current Thread [Next in Thread]