[Bug fortran/59198] [4.7/4.8/4.9 Regression] ICE on cyclically dependent polymorphic types

[janus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59198

--- Comment #5 from janus at gcc dot gnu.org ---
The ICE and backtrace one gets on all variants of the test case has changed
slightly to:


internal compiler error: in tree_to_uhwi, at tree.h:3653
      class(rng_t), allocatable :: rng
 ^
0xb24434 tree_to_uhwi
    /home/jweil/gcc49/trunk/gcc/tree.h:3653
0xb24434 output_constructor_regular_field
    /home/jweil/gcc49/trunk/gcc/varasm.c:4894
0xb24434 output_constructor
    /home/jweil/gcc49/trunk/gcc/varasm.c:5231
0xb237aa assemble_variable(tree_node*, int, int, int)
    /home/jweil/gcc49/trunk/gcc/varasm.c:2139
0xb25285 varpool_assemble_decl(varpool_node*)
    /home/jweil/gcc49/trunk/gcc/varpool.c:455
0x6d0b90 output_in_order
    /home/jweil/gcc49/trunk/gcc/cgraphunit.c:2010
0x6d0b90 compile()
    /home/jweil/gcc49/trunk/gcc/cgraphunit.c:2247
0x6d0e59 finalize_compilation_unit()
    /home/jweil/gcc49/trunk/gcc/cgraphunit.c:2329
0x8396b4 write_global_declarations()
    /home/jweil/gcc49/trunk/gcc/langhooks.c:323



Not sure if this is actually a Frotran problem or rather originates from the
middle-end ...