Dear all, Jose posted a patch here that was never reviewed:
https://gcc.gnu.org/pipermail/fortran/2021-April/055933.html I could not find any issues with his patch, it works as advertised and fixes the reported problem. As his testcases did not reliably fail without the patch but rather randomly due to the uninitialized descriptor, I added a check of the tree-dumps to verify that the TKR initializer is generated. Does anybody else have any comments? Regtested on x86_64-pc-linux-gnu. OK for mainline? Thanks, Harald