gfortran-8 compilation error

[Jonas Wittbrodt]

Hey MG5 team,

Madgraph (MG5_aMC-2.6.3.2) does not work with gfortran-8.1. I get the following compilation error after launch:

genps.f:60:55:
    call gen_mom(iconfig,mincfig,maxcfig,invar,wgt,x,p)
                                              1
Error: Actual argument contains too few elements for dummy argument ‘p1’ (5/20) at (1)

This happens for any process and model I tried (including SM e+ e- > mu+ mu- for which the log file is attached). I imagine the -fbounds-check got stricter with gfortran-8 and now catches this.

I will go back to gfortran-5 for now but would appreciate if you could fix this.

Cheers and thanks,
Jonas

[Ivan Razumov (iarspider)]

Hi Jonas,

we at GENSER have also encountered this problem. You can use the patch (attached) to fix it.

Best regards,
Ivan

[Olivier Mattelaer (olivier-mattelaer)]

Hi,

The development version (2.6.4) is currently tagged as compatible for LO run with gcc8.
This being said, NLO is not yet supported by gcc8 and
It is likely that some feature will not be fully working in 2.6.4 with gcc8.
We will need help of all users to find all the issues related to gcc8 and I do not expect this to be fully stable before a while.

Cheers,

Olivier

> On 2 Jul 2018, at 11:31, Jonas Wittbrodt <email address hidden> wrote:
>
> Public bug reported:
>
> Hey MG5 team,
>
> Madgraph (MG5_aMC-2.6.3.2) does not work with gfortran-8.1. I get the
> following compilation error after launch:
>
> genps.f:60:55:
> call gen_mom(iconfig,mincfig,maxcfig,invar,wgt,x,p)
> 1
> Error: Actual argument contains too few elements for dummy argument ‘p1’ (5/20) at (1)
>
> This happens for any process and model I tried (including SM e+ e- > mu+
> mu- for which the log file is attached). I imagine the -fbounds-check
> got stricter with gfortran-8 and now catches this.
>
> I will go back to gfortran-5 for now but would appreciate if you could
> fix this.
>
> Cheers and thanks,
> Jonas
>
> ** Affects: mg5amcnlo
> Importance: Undecided
> Status: New
>
> ** Attachment added: "log file"
> https://bugs.launchpad.net/bugs/1779636/+attachment/5158590/+files/run_01_tag_1_debug.log
>
> --
> You received this bug notification because you are subscribed to
> MadGraph5_aMC@NLO.
> https://bugs.launchpad.net/bugs/1779636
>
> Title:
> gfortran-8 compilation error
>
> Status in MadGraph5_aMC@NLO:
> New
>
> Bug description:
> Hey MG5 team,
>
> Madgraph (MG5_aMC-2.6.3.2) does not work with gfortran-8.1. I get the
> following compilation error after launch:
>
> genps.f:60:55:
> call gen_mom(iconfig,mincfig,maxcfig,invar,wgt,x,p)
> 1
> Error: Actual argument contains too few elements for dummy argument ‘p1’ (5/20) at (1)
>
> This happens for any process and model I tried (including SM e+ e- >
> mu+ mu- for which the log file is attached). I imagine the -fbounds-
> check got stricter with gfortran-8 and now catches this.
>
> I will go back to gfortran-5 for now but would appreciate if you could
> fix this.
>
> Cheers and thanks,
> Jonas
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/mg5amcnlo/+bug/1779636/+subscriptions

[Jonas Wittbrodt (nandurius)]

Hi Olivier,

thanks for the reply. Great to hear that you are already working on it. I will gladly try out 2.6.4 with gcc8 once it is available.

Also thanks to Ivan for the patch suggestion, setting -std=legacy should do very similar things to going back to the older compiler.

Cheers,
Jonas