Vsd value set of pointers #5825
Conversation
Codecov Report
@@ Coverage Diff @@
## develop #5825 +/- ##
===========================================
+ Coverage 72.85% 72.90% +0.05%
===========================================
Files 1421 1425 +4
Lines 154130 154236 +106
===========================================
+ Hits 112292 112452 +160
+ Misses 41838 41784 -54
Continue to review full report at Codecov.
|
jezhiggins
force-pushed
the
vsd-value-set-of-pointers
branch
2 times, most recently
from
85160cc to
8342a92
Compare
jezhiggins
requested review from
chrisr-diffblue,
martin-cs,
peterschrammel and
smowton
as code owners
| @@ -2,23 +2,20 @@ | |||
|
|
|||
| Module: analyses variable-sensitivity | |||
|
|
|||
| Author: Thomas Kiley, [email protected] | |||
| Author: Jez Higgins, [email protected] | |||
There was a problem hiding this comment.
Although the author fields are a bit meaningless in the age of git praise / git blame, we do prefer to have the name of the original author.
There was a problem hiding this comment.
I have a feeling this might be something of a git artefact, but happy to make the changes.
There was a problem hiding this comment.
If these are new files (that git has just decided are moved) then they should have you as the author. If they are moved they should keep the original author.
There was a problem hiding this comment.
Honestly, I can't remember, and as you say not hugely important. However, since it obviously looked odd in the diff, I've made the changes.
| @@ -2,84 +2,52 @@ | |||
|
|
|||
| Module: analyses variable-sensitivity | |||
|
|
|||
| Author: Thomas Kiley, [email protected] | |||
| const exprt &expr, | ||
| const abstract_environmentt &environment, | ||
| const namespacet &ns); | ||
|
|
||
| /** |
There was a problem hiding this comment.
If the function stays then it would be good if the doxygen stays / gets updated.
|
|
||
| Author: Diffblue Ltd. |
Introduce new base class, abstract_pointer_objectt. It should be, but is not yet, an abstract (in C++ terms) class as it's not intended to be directly instantiable. pointer_abstract_object now renamed two_value_pointer_abstract_objectt. constant_pointer_abstract_objectt and two_value_pointer_abstract_objectt are now peers. Deleted the unused value_set_pointer_abstract_objectt in anticipation of reintroducing a more functional value-set pointer representation shortly.
Stripped back value_set_abstract_object now it doesn't have to deal with pointers. Pulled out abstract_object_sett as a type not just an alias. Add behaviour with functions pulled across from value_set_abstract_objectt and value_set_pointer_abstract_objectt.
Remove expression_transform and write from value-set of pointers. Dereference value-set of pointers to a single value
Test uses intervals. Value-sets should also work but we can't write that test yet as we needs more work on asserting value-set equality.
Tests determinate and indeterminate write through a pointer
jezhiggins
force-pushed
the
vsd-value-set-of-pointers
branch
from
8342a92 to
df41607
Compare
This leads on from #5793, separating out value-sets of constants from value-sets of pointers.
Prior to this work,
value_set_abstract_objecttdid double duty for constants and pointers, contained a certain amount of runtime type checking to do the right thing, doing a decent job for values, but a poor job for pointers. Separating the two out isolates the constant and pointer specific behaviours, immediately increasing clarity and enabling some easy fixes for pointer dereferencing and writing through.I could have gone further, as the pointer behaviour, while improved, is not complete, and there's still a certain amount of duplication that could be eliminated. However, I believe much of this will be addressed, and actually more easily addressed, in the upcoming working on multiple-dispatch/cross-representation evaluation and merge.