https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118747
--- Comment #1 from anlauf at gcc dot gnu.org --- Not sure what kind of regression this is. The situation is "interesting". Replacing deferrred-length by fixed-length leads to a crash with gcc-14, too. (e.g. when using character(len=3), allocatable :: string_) Compiling with deferrred-length, but with optimization, hides the problem somehow, at least with the testcase. The tree-dump suggests that we double-free, but I have some doubt about the code generated by gcc-14, too. There is free'ing going on in the elemental loop on the one side, while gcc-15 does a free of the string when destructing the constructor. I don't see at the moment what the right intention is with the temporary for the constructor.
