https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121411
--- Comment #5 from David Faust <david.faust at oracle dot com> --- Yeah, looks like there are multiple inconsistencies with uint32/unsigned int versus unsigned HOST_WIDE_INT when fetching values from the DWARF DIEs. The issues with arrays is one, bit_size in gen_ctf_sou_type is another. ctf_die_bitsize should also be returning unsigned HWI. It seems to me that an underlying problem and likely source of mistakes is that add_AT_unsigned accepts and stores uHWI values, but get_AT_unsigned returns unsigned (int) not uHWI. This asymmetry could be causing silent truncation anywhere; HWI is supposed to always be 64-bits. There are uses in dwarf2out I am suspicious of now as well, DW_AT_byte_size is stored as uHWI and elsewhere read as unsigned int...
