https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109189
--- Comment #1 from Jan Dubiec <jdx at o2 dot pl> --- BTW, it would be nice if someone experienced could inspect lines 1547–1553, 1572–1578 and 1751–1757 in that file. These ranges contain code which looks like this: case CONST_DOUBLE: { long val; REAL_VALUE_TO_TARGET_SINGLE (*CONST_DOUBLE_REAL_VALUE (x), val); fprintf (file, "#%ld", ((val >> 16) & 0xffff)); break; } They seem kinda suspicious to me.