[clang] [analyzer] Fix false positive for bitfield read after cast-pointer write (PR #188387)

[Balázs Benics via cfe-commits]

================
@@ -2554,6 +2638,11 @@ RegionStoreManager::bind(LimitedRegionBindingsConstRef B, Loc L, SVal V) {
   // Clear out bindings that may overlap with this binding.
   auto NewB = removeSubRegionBindings(B, cast<SubRegion>(R));
 
+  // When writing a scalar through a cast pointer to a record region
+  // (e.g. *(unsigned*)&struct_var = val), track which bits of each
+  // overlapping bitfield have been written using a per-field bitmask.
+  NewB = updateBitfieldCoverage(B, NewB, R, V);
+
----------------
steakhal wrote:

One reason for why we should not change how `bind` works is because of the following invariant:
If we store something somewhere, and then right after that we load it back; we should get the value we stored there.
This applies to writes and stores to unaligned, type-punned storage.
This means that the value we store, should not care if the memory beneath is a bitfield or anything else - given that there is actually some memory.

https://github.com/llvm/llvm-project/pull/188387