https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87415
Aldy Hernandez <aldyh at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Assignee|unassigned at gcc dot gnu.org |aldyh at gcc dot gnu.org
--- Comment #2 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
All mine baby!
For a 1-bit signed field we are trying to subtract the following ranges:
[0, 0] - VR_VARYING
Mathematically these are:
[0, 0] - [MIN, MAX]
=> [0, 0] - [-1, 0]
[Because 1-bit signed fields go from -1 to 0, supposedly to make two's
complement work, but really to piss me off.]
For ranges: [a, b] - [c, d] is [a - d, b - c], so combine_bounds() yields:
[0, OVERFLOW]
Then we adjust the result for overflows in set_value_range_with_overflow() we
transform the above into:
[0, +INF]
And since +INF is 0, we get [0, 0] which is incorrect.
[0, 0] - [MIN, MAX] should have given [MIN, MAX].
I'm still trying to work this out without special casing 1-bit fields. But the
original patch is correct-- perhaps we should adjust
set_value_range_with_overflow... not sure yet.
