Update Github runners to Ubuntu 24.04

[samhatfield]

No description provided.

[samhatfield]

Note that the ectrans_test_adjoint test is broken for intel-llvm (ifx) due to a compiler bug. The following program behaves wrong when compiling with ifx and -fiopenmp:

PROGRAM REPRODUCER

IMPLICIT NONE

INTEGER, PARAMETER :: N = 5
INTEGER :: ARRAY(N)

ARRAY(:) = 1

PRINT *, "Values in PROGRAM"
PRINT *, ARRAY(1:N)
CALL TEST1

CALL TEST2(ARRAY)

CONTAINS

  SUBROUTINE TEST1
    PRINT *, "Values in TEST1"
    PRINT *, ARRAY(1:N)
  END SUBROUTINE TEST1

  SUBROUTINE TEST2(KVSET)
    INTEGER, INTENT(IN) :: KVSET(:)

    PRINT *, "Values in TEST2"
    PRINT *, KVSET
  END SUBROUTINE TEST2

END PROGRAM REPRODUCER

I've reported this to Jacob Poulsen at Intel.

[samhatfield]

Good to merge.

[wdeconinck]

I would add the ifx compiler bug in a separate PR for visibility. This PR is then to be merged afterwards.

[wdeconinck]

ectrans_test_install fails??? That is very concerning...

[samhatfield]

Yeah, I'll reenable it so you can see the error.

[samhatfield]

See https://github.com/ecmwf-ifs/ectrans/actions/runs/13815938473/job/38649010125?pr=224#step:12:4336. Tricky to debug non-interactively.

[wdeconinck]

It seems that only the Fortran executables are working and the C executables transi_sptogp or transi_gptosp are causing a segfault. I wonder if we can try to reproduce this locally without a GitHub-runner.

[samhatfield]

I had a go on Atos with the same version of Intel (2023.2.0) but the test passed. I think we'll need to debug this through the action itself.

[samhatfield]

See https://github.com/ecmwf-ifs/ectrans/pull/227/files for ifx bug fix.

[wdeconinck]

See https://github.com/ecmwf-ifs/ectrans/pull/227/files for ifx bug fix.

OK that is merged now. This can be rebased.