| commit | e678694e3a107c9689b104d85cfad637400cc4cd | |
| author | Dan Carpenter <dan.carpenter@oracle.com> | |
| Fri, 9 Nov 2012 09:26:37 +0000 (9 12:26 +0300) | ||
| committer | Dan Carpenter <dan.carpenter@oracle.com> | |
| Fri, 9 Nov 2012 09:26:37 +0000 (9 12:26 +0300) | ||
| tree | 274b83b0a7371816537b2d4a43920f5e4b29c161 | treesnapshot (tar.gz zip) |
| parent | f3523565e83414e5cb08eca4de0f1b29adfb5f47 | commitdiff |
sval: fuzzy_max: fix it to work with svals better
Fuzzy max was using llong_ctype instead of the type of the estate. So if
the comparisons were promoted to unsigned the min became a very large
number.
Also I introduced the is_some_kind_of_max() function. In the olden days
we could just compare against whole_range.max but these days that is gone.
We do try to store things in smatch_absolute, but that doesn't work always
because what if we know the state is non-zero? The smatch_implied
information has to save that it is "1-u32max" or whatever.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Fuzzy max was using llong_ctype instead of the type of the estate. So if
the comparisons were promoted to unsigned the min became a very large
number.
Also I introduced the is_some_kind_of_max() function. In the olden days
we could just compare against whole_range.max but these days that is gone.
We do try to store things in smatch_absolute, but that doesn't work always
because what if we know the state is non-zero? The smatch_implied
information has to save that it is "1-u32max" or whatever.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
| smatch_math.c | diffblobblamehistory |