PDP-11: _dbl_ext_cvt() does not work properly

[kwhr0]

On PDP-11, integer and floating point number conversion instructions are signed.
In _dbl_ext_cvt() of lang/cem/libcc.ansi/core/stdlib/ext_comp.c, the return value of frexp() is 0.5<=value<1.0.

value *= 4294967296.0;
e->m1 = value;

Converting as above will overflow with signed 32 bits, so it will not work correctly.
If you replace the same function with the one in lang/cem/libcc/gen/ext_comp.c, it will work correctly.