================
@@ -108,6 +108,16 @@ class DataflowAnalysisContext {
/// A null `PointeeType` can be used for the pointee of `std::nullptr_t`.
PointerValue &getOrCreateNullPointerValue(QualType PointeeType);
+ /// Adds `Constraint` to current and future flow conditions in this context.
+ ///
+ /// The global condition must contain only flow-insensitive information, i.e.
----------------
sam-mccall wrote:
This is a much more accurate descriptor, but IME "flow insensitive" is not good
as an actual name. We used "non-flow-sensitive" extensively in nullability and
gradually phasing it out in favor of the "type" metaphor, as after many months
reading it still requires "deciphering". Mostly it's too long, it's a negative
definition, and too technical I think.
The parallel between `addFlowConditionConstraint` and
`addFlowInsensitiveInvariant` is also IMO not clear enough. (I'd be in favor of
dropping "Constraint" from both, but that's a different change)
I've tried `Common` instead of `Global` which might more clearly emphasize the
sharing that happens here - WDYT?
If you feel strongly about `FlowInsensitive` I can use that, it might be more
consistent with the framework's style.
https://github.com/llvm/llvm-project/pull/65949
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
