vamp numerical failure

[Juergen Reuter]

This is a spurious error, that is getting on our nerves for tooooo long...

[Juergen Reuter]

This is the error message:

Runtime Error: *** Arithmetic exception: Floating invalid operation - aborting
../../../../src/vamp/src/vamp.f90, line 1551: Error occurred in VAMP_REST:VAMP_PRINT_ONE_HISTORY
../../../../src/vamp/src/vamp.f90, line 1591: Called by VAMP_REST:VAMP_PRINT_HISTORIES
../../../../src/vamp/tests/vamp_test.f90, line 284: Called by VAMP_TESTS:MULTI_CHANNEL
../../../../src/vamp/tests/vamp_test.f90, line 365: Called by VAMP_TEST

[Juergen Reuter]

I did a screening over that test, this numerical failure happened in 2 out of 100 tests. I will try to trace it further down, valgrind did not find anything, as it is no memory leak or so. More later...

[Juergen Reuter]

Guess I have that, doing a cycle over 500 tests now to check whether I cured the problem, and then that Heisenbug is no longer bothering us...hopefully

[Juergen Reuter]

There is another uncaught floating exception, this times it is not only in the print procedures:

Runtime Error: *** Arithmetic exception: Floating invalid operation - aborting
coordinates.f90, line 173: Error occurred in COORDINATES:CARTESIAN_TO_SPHERICAL_COS_2
coordinates.f90, line 186: Called by COORDINATES:CARTESIAN_TO_SPHERICAL_COS
vamp_test.f90, line 125: Called by VAMP_TEST_FUNCTIONS:IHP
vamp.f90, line 1742: Called by VAMP_REST:VAMP_MULTI_CHANNEL
vamp_test.f90, line 164: Called by VAMP_TEST_FUNCTIONS:W
vamp.f90, line 877: Called by VAMP_REST:VAMP_SAMPLE_GRID0
vamp.f90, line 2051: Called by VAMP_REST:VAMP_SAMPLE_GRIDS
vamp_test.f90, line 288: Called by VAMP_TESTS:MULTI_CHANNEL
vamp_test.f90, line 365: Called by VAMP_TEST

Will try to get to that one.

[Juergen Reuter]

Was not so difficult to catch that one, but the code was used only for the test routine, not for the integrations themselves. Running an extensive scan now, when it does not crash any more, I will check it in and close the ticket.

[Juergen Reuter]

Uff, there is yet another one. Catching that, too, starting another scan test.

[Juergen Reuter]

Done in r3651. No exception any more in 500 checks each.