================
@@ -14,18 +14,52 @@
#define DEBUG_TYPE "dataflow"
-void clang::dataflow::copyRecord(RecordStorageLocation &Src,
- RecordStorageLocation &Dst, Environment &Env)
{
+namespace clang::dataflow {
+
+static void copyField(const ValueDecl *Field, StorageLocation *SrcFieldLoc,
----------------
martinboehme wrote:
> Could `SrcFieldLoc` be a ptr to const? (Same below).
I think what you want to express is that the source field location is only ever
read from, while the destination field location is written to?
This isn't really how we use `StorageLocation` constness in the framework
today. (Maybe we should -- that's a separate discussion -- though I'm not
convinced about how much this would buy us.) Grepping for `const
StorageLocation` yields only a few hits in the whole framework[^1], and I think
that's for good reason: A `StorageLocation` is essentially immutable anyway;
all of its method are const, as are almost all of the methods in the derived
class `RecordStorageLocation`, with the exception of `setChild()` (which is a
bit of a special case). It's similar in this respect to a
[flyweight](https://en.wikipedia.org/wiki/Flyweight_pattern). So the
distinction between `const StorageLocation *` and `StorageLocation *` doesn't
mean much in practice, and I guess we've gravitated to the less verbose
`StorageLocation`.
We _could_ consider using `StorageLocation` constness to encode "can this
storage location be written to" (which I think is your intent?), which would
entail changing the storage location parameter of `setValue(const
StorageLocation &, Value &)` to be non-const, but that would be a pretty
far-reaching change because I think we'd need to add const in a lot of places
in the framework.
[^1]: Notably, `setValue(const StorageLocation &, Value &)` is one of them, so
passing a const storage location does not prevent us from writing to that
storage location.
https://github.com/llvm/llvm-project/pull/85064
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
