test suite fails with nagfor

[kilian]

Several of the standard tests fail when the NAG compiler is used. These are some of the self-checks that we recently re-enabled without properly checking what they were doing. This has to be investigated. With gfortran, no failures.

[Juergen Reuter]

Duplicate of #235.

[kilian]

No, this was without -nan. Compiler flags: -C=all -gline -f2003 -O2. The failed scripts are:

FAIL: particles.run                                                             
FAIL: polarizations.run                                                         
FAIL: hepmc.run                                                                 

Appears to be unrelated to #235.

[Juergen Reuter]

Oh well, yes, I know, this comes becomes Jenkins is omitting the C=all flag. After I've fixed this with this we should switch this on.... Please, don't confuse me.... :(

[Juergen Reuter]

The dangling pointer was caused by a missing target specification in polarization_get_axis. Now, there is only one failing test, particles, because of a dimensional mismatch.

[Juergen Reuter]

The dimensional mismatch comes from the following: at the end of the particle test the particle_set3 is written raw to a unit, the unit rewinded, then read raw from the unit into particle_set4, and this then wrote via particle_set_write. This fails due to the mismatch in prt_set%pol%state%me .

[Juergen Reuter]

It seems that the raw version do not write the state%me components which are then missing. WK's comment: Binary I/O. Matrix elements are written only for leaf nodes. this was actually a pointer is lost. First, grammatically not correct, secondly, I believe there is something foul here though not necessarily in Denmark. Disabling the test, closing the ticket. I believe that there is still an issue, but its not dramatic, and it is not used anywhere.

[kilian]

Did you really check your changes in? On my machine, there are still the same three tests failing with nagfor -C=all, and on Jenkins as well.

[Juergen Reuter]

The culprit was the accuser himself: by an unclean commit (however that happened) WK has erased my changes... "ubergrmpf*