fix(labels): decode non-ASCII identifiers in Rust v0 symbols

src/smda/common/labelprovider/rust_demangler/rust_v0.py

@@ -101,7 +101,7 @@ def insert(self, i: int, c: str) -> bool:
         self.out[i] = c
         return True
 
-    def punycode_decode(self) -> Optional[None]:
+    def punycode_decode(self) -> Optional[bool]:
         count = 0
         punycode_bytes = self.punycode
         try:
@@ -151,10 +151,12 @@ def punycode_decode(self) -> Optional[None]:
             n += i // lent
             i %= lent
 
-            try:
-                c = chr(n)
-            except (ValueError, OverflowError):
+            # mirror char::from_u32: a scalar value, so no surrogate and nothing past the
+            # last code point. chr() accepts surrogates, and a name carrying one cannot be
+            # encoded as UTF-8 by whoever consumes the report.
+            if not 0 <= n <= 0x10FFFF or 0xD800 <= n <= 0xDFFF:
                 return None
+            c = chr(n)
 
             if not self.insert(i, c):
                 return None
@@ -163,7 +165,7 @@ def punycode_decode(self) -> Optional[None]:
             try:
                 punycode_bytes[count]
             except IndexError:
-                return
+                return True
 
             delta = delta // damp
             damp = 2

tests/testRustSymbolProvider.py

@@ -11,6 +11,7 @@
 from smda.common.labelprovider.rust_demangler.rust import TypeNotFoundError
 from smda.common.labelprovider.rust_demangler.rust_legacy import LegacyDemangler, UnableToLegacyDemangle
 from smda.common.labelprovider.rust_demangler.rust_v0 import (
+    Ident,
     Parser,
     Printer,
     UnableTov0Demangle,
@@ -113,6 +114,28 @@ def test_v0_unnamed_closure_has_no_stray_colon(self):
         # unnamed closures/shims must render as {closure#N}, not {closure:#N}
         self.assertEqual(demangle("_RNCNvC8rustc_v01fs_0"), "rustc_v0::f::{closure#1}")
 
+    def test_v0_non_ascii_identifier_is_decoded(self):
+        # the decode itself always worked; its success was reported as None,
+        # which is what every failure path returns, so the caller fell back
+        name = "_RNqCs4fqI2P2rA04_11utf8_identsu30____7hkackfecea1cbdathfdh9hlq6y"
+        self.assertEqual(demangle(name), "utf8_idents::საჭმელად_გემრიელი_სადილი")
+        self.assertNotIn("punycode{", demangle(name))
+
+    def test_v0_punycode_never_decodes_to_a_surrogate(self):
+        # chr() accepts 0xD800-0xDFFF where Rust's char::from_u32 refuses; a name carrying
+        # one cannot be encoded as UTF-8 by whoever consumes the report
+        name = "_RNvCu4_ib9b4main"
+
+        demangled = demangle(name)
+
+        self.assertNotIn("\ud800", demangled)
+        demangled.encode("utf-8")
+
+    def test_v0_undecodable_punycode_still_falls_back_to_the_raw_form(self):
+        ident = Ident("", "!not-punycode!")
+        ident.display()
+        self.assertEqual(ident.disp, "punycode{!not-punycode!}")
+
     def test_v0_empty_const_hex_nibbles_raise_demangler_error(self):
         # `Kh_` (u8 const with zero hex nibbles) previously escaped as a bare
         # ValueError from int("", 16) instead of the demangler's own error type

tests/test_fuzz_rust_demangler.py

@@ -26,6 +26,9 @@ def test_rust_demangler_never_hangs(s):
     """
     try:
         result = demangle(s)
+        # a demangled name is handed to consumers as text; a lone surrogate would make it
+        # unencodable, so the demangler must never produce one
+        result.encode("utf-8")
     except (TypeNotFoundError, UnableTov0Demangle, UnableToLegacyDemangle):
         return
     except RecursionError: