On 11/13/25 1:56 PM, Harald Anlauf wrote:
Dear all,
this patch addresses the issue that a procedure may appear in
a derived-type constructor as target of a procedure pointer.
Adjusting check_constructor solves the rejects-valid.
While working on an extended testcase, I found that there is a
resolution issue in gfortran when trying to use dispatch tables
that are named constants (i.e., PARAMETER). I have not yet been
able to locate where the difference between dynamic and static
tables leads to different resolution, and decided to add all
variants to the testcase with the wrongly rejected cases
commented out. Note that the full testcase works with NAG.
I propose to commit the attached variant and leave the PR open
until the resolution issue is fixed or the PR is replaced by
a more appropriate one.
Regtested on x86_64-pc-linux-gnu. OK for mainline?
Thanks,
Harald
All good, OK for mainline.
Jerry
