On 01.02.21 19:28, Paul Richard Thomas wrote:
I have attached a memory leak free version of the testcase. I have asked for Thomas's help to use frontend-passes.c tools to do the same for compound constructors with allocatable components. My attempts to do the job in other ways have failed totally.
Well, if we can do it in the FE passes, why not. I also do not mind if we have memory leaks in the testcase – if either the standard permits it or if (as here) a PR exists, which tracks the issue. (Especially as it is not a new issue.) I am fine with either testcase – the non-memory-leaking one is nicer, if we are reasonably sure it tests the right thing. (I think we are.) If not, we should instead/additionally add the currently leaking variant. Thanks for looking into this, Tobias ----------------- Mentor Graphics (Deutschland) GmbH, Arnulfstraße 201, 80634 München / Germany Registergericht München HRB 106955, Geschäftsführer: Thomas Heurung, Alexander Walter
