On 6/27/25 1:08 PM, Toon Moene wrote:
On 6/27/25 17:41, Toon Moene wrote:
Note: as the timeline at the top of the source implies, I have worked on this
until early 2021.
This means it compiled with gfortran 10 and 11.
So the problem is probably introduced after these versions.
[ Of course, while I perused the Fortran 2018 standard while writing this
code, I could easily have made an error that gfortran 10/11 simply didn't
catch. ]
I just used the following:
--- snip ---
Using latest 14 it does compile fine with -fcoarray=single.
I have not found the equivalent of the libcoarrays-openmpi-dev
package yet on Fedora 42. As soon as I hunt that down I will see if I can
reproduce your run.
In the meantime this confirms we have a regression on 15 and 16.
Very helpful,
Jerry
