fix(test): pointer identity for a 'static name is not assertable on MSVC

changelog.d/8239-windows-static-literal-identity.md

@@ -0,0 +1,16 @@
+`windows-build` is green again, restoring the whole `test.yml` run that
+`release-packages`' `await-tests` requires before it will cut a tag.
+
+Two tests from #8177 failed on Windows and only there, asserting that a bound
+method's closure captures the `'static` method-name literal by comparing
+pointers. Both failing pairs differed by the same constant offset (0x161F90) —
+two copies of the same read-only data, not a heap pointer. ELF
+(`SHF_MERGE|SHF_STRINGS`) and Mach-O (`__TEXT,__cstring`) merge identical
+read-only strings, so the copy the closure captures and the copy the lookup
+returns land at one address; MSVC does not pool identical literals across
+codegen units, so they stay distinct. Both are `'static`.
+
+The address comparison is now gated on a linker that merges. The property the
+test exists for — the captured name must not be the movable key string's
+interior — is asserted unconditionally, as are the captured length and bytes, so
+Windows keeps the coverage and loses only the proxy it cannot evaluate.

crates/perry-runtime/src/gc/tests/handle_bound_method_name.rs

@@ -84,6 +84,20 @@ unsafe fn assert_names_the_literal(
     what: &str,
 ) {
     let (name_ptr, name_len) = captured_name(bound);
+    // Pointer identity is only assertable where the linker merges identical
+    // read-only strings. ELF (`SHF_MERGE|SHF_STRINGS`) and Mach-O
+    // (`__TEXT,__cstring`) do; MSVC does not pool identical literals across
+    // codegen units, so the copy the closure captures and the copy the lookup
+    // returns can be two distinct `&'static [u8]` at different addresses.
+    // Measured on the Windows runner: both failing pairs differed by the SAME
+    // constant offset (0x161F90), i.e. two whole copies of the same read-only
+    // data, not a heap pointer.
+    //
+    // Both are `'static`, which is the property this test exists for: the name
+    // must not be the MOVABLE key string's interior. That invariant is asserted
+    // unconditionally below, together with the length and the bytes, so Windows
+    // keeps real coverage — it just cannot use address equality as the proxy.
+    #[cfg(not(windows))]
     assert_eq!(
         name_ptr,
         expected.as_ptr(),