[Patch] OpenMP/Fortran: Fix handling of optional is_device_ptr + bind(C) [PR108546]

(That's a[11/12/13 Regression] P2 regression) The problem is that an is-present check on the receiver side (inside the target region) does not make much sense; the != NULL check needs to be done before the GOMP_target_ext but it *is* already there. (Having the check inside the target region failed

Re: [Patch] OpenMP: Handle descriptors in target's firstprivate [PR104949]

Hi! I'm currently reviewing 'gomp_copy_host2dev', 'ephemeral' in a different context, and a question came up here; commit r13-706-g49d1a2f91325fa8cc011149e27e5093a988b3a49 "OpenMP: Handle descriptors in target's firstprivate [PR104949]": On 2022-05-11T19:33:00+0200, Tobias Burnus wrote: > this p