Edit**3: A patch has been attached to the bugzilla report. It is sufficient that for most users gfortran will do what you want. I no longer commit patches to the actually GCC source code repository. So, if someone wants to champion a patch and get involved with gfortran developement, here’s your opportunity.
I did implement the F2018 feature. The patch attached to the bug report has been superseded by one in my local source code repository. import none and import all work. import, only : i works with a caveat. The caveat is that how import was originally implemented needs to be ripped out and redone. I doubt I’ll submit a new patch. A correct fix would require someone sponsoring the work. import none is the only case of interest to me.
Do you think you could push that patch at some point to be merged to trunk? One of the things that has stopped be from properly implementing import host associations in fortls was the fact that gfortran did not support them. It would be awesome if the patch got included with the next release.
At this point in time, I will not be submitting the patch. Checking the constraints, import, none, and import, all appear to work. Unfortunately, import, only : works for a very limited situation, which renders this usage as unusable.
This is an example, where J3 made a change of a feature in the language, that looks great from a user viewpoint. It unfortunately has the side effect that a vendor may need to redesign how import was implemented to meet F2003 and F2008 requirements. This is fine for a commercial vendor, who has a financial incentive to fix its compiler. For a project, which is primarily supported by volunteers, it becomes a burden.
Unfortunately, I understand very well where you are coming from. It is truly unfortunate that gfortran is not better funded. Judging from the results in the Fortran Developer's Poll - 2022gfortran is almost universally used by all the participants at some point in the development/deployment cycle.
I am certain that there exist financially stable and profitable companies/institutions that use gfortran but have not considered donating the the gcc project.