| commit | c9e5f91d4e3ee48d850848a522af5fdf95de15a0 | |
| author | Dan Carpenter <error27@gmail.com> | |
| Wed, 18 Feb 2009 08:42:32 +0000 (18 11:42 +0300) | ||
| committer | Dan Carpenter <error27@gmail.com> | |
| Wed, 18 Feb 2009 08:42:32 +0000 (18 11:42 +0300) | ||
| tree | ed695567825abc655762ca8f517dfebb0e302b49 | treesnapshot (tar.gz zip) |
| parent | 7fdb8f8941c6c58d7f000b04640b565cc8b19184 | commitdiff |
fixes to the -> possible setting.
The new rule is that ->possible should never be NULL. It should always
at least point to itself.
When you "or" the states the ->possible list was not getting updated.
When you merge a state with a NULL path then that adds an &undefined to
the ->possible list. This is important for "if (a || b)". In that case
you merge a &nonnull with a NULL which means the possible states are
&nonnull, &merged, and &undefined.
The second fix was that clone_state() and as result overwrite_sm_state()
were not saving the ->possible list.
Signed-off-by: Dan Carpenter <error27@gmail.com>
The new rule is that ->possible should never be NULL. It should always
at least point to itself.
When you "or" the states the ->possible list was not getting updated.
When you merge a state with a NULL path then that adds an &undefined to
the ->possible list. This is important for "if (a || b)". In that case
you merge a &nonnull with a NULL which means the possible states are
&nonnull, &merged, and &undefined.
The second fix was that clone_state() and as result overwrite_sm_state()
were not saving the ->possible list.
Signed-off-by: Dan Carpenter <error27@gmail.com>
| smatch_slist.c | diffblobblamehistory |