Handle bool expressions in implicit_typecast_arithmetic #5758
Conversation
Codecov Report
@@ Coverage Diff @@
## develop #5758 +/- ##
=========================================
Coverage 76.73% 76.74%
=========================================
Files 1579 1579
Lines 182008 182164 +156
=========================================
+ Hits 139671 139795 +124
- Misses 42337 42369 +32
Continue to review full report at Codecov.
|
src/ansi-c/c_typecast.cpp
Outdated
| @@ -619,8 +619,10 @@ void c_typecastt::implicit_typecast_arithmetic( | |||
| if(max_type==LARGE_SIGNED_INT || max_type==LARGE_UNSIGNED_INT) | |||
| { | |||
| // get the biggest width of both | |||
| std::size_t width1 = to_bitvector_type(type1).get_width(); | |||
| std::size_t width2 = to_bitvector_type(type2).get_width(); | |||
| std::size_t width1 = | |||
There was a problem hiding this comment.
Would it make sense to use c_type1 and c_type2 instead of type1 and type2?
There was a problem hiding this comment.
Agreed; this should be post-promotion.
There was a problem hiding this comment.
But then we'd need new facilities to turn a c_typet into a typet. It's kind-of done by c_typecastt::implicit_typecast_arithmetic, but that doesn't do anything about a LARGE_SIGNED_INT (or LARGE_UNSIGNED_INT).
There was a problem hiding this comment.
I've now implemented this without a need for to_bitvector_type at all. Maybe I'm missing something? The new implementation seems almost too simple.
src/ansi-c/c_typecast.cpp
Outdated
| @@ -619,8 +619,10 @@ void c_typecastt::implicit_typecast_arithmetic( | |||
| if(max_type==LARGE_SIGNED_INT || max_type==LARGE_UNSIGNED_INT) | |||
| { | |||
| // get the biggest width of both | |||
| std::size_t width1 = to_bitvector_type(type1).get_width(); | |||
| std::size_t width2 = to_bitvector_type(type2).get_width(); | |||
| std::size_t width1 = | |||
There was a problem hiding this comment.
Agreed; this should be post-promotion.
We use Booleans instead of int as type of binary predicates, and enums don't convert to bitvector types directly either. This must not preclude mixing binary predicates with bitvector-typed expressions. Fixes: diffblue#5103
We use Booleans instead of int as type of binary predicates. This must
not preclude mixing binary predicates with bitvector-typed expressions.
Fixes: #5103