Address review: abort on OOM, keep _extra_ops out of librt build

- codepoint_extra_ops.h: include CPy.h and move the isidentifier slow
  path inline into LibRTStrings_IsIdentifier. Aborts via
  CPyError_OutOfMemory on allocation failure (the helper is ERR_NEVER,
  so returning a silently-wrong bool under memory pressure was the wrong
  contract). Matches the pattern in the sibling _extra_ops.h files (all
  bodies static inline, CPy.h included for runtime helpers).
- codepoint_extra_ops.c: reduce to a single-line shim that #includes the
  header. Exists only so SourceDep("codepoint_extra_ops.c") pulls the
  header into user mypyc extensions in include_runtime_files mode.
- build.py / lib-rt/setup.py: drop codepoint_extra_ops.c from the
  librt.strings module sources. The _extra_ops.c files are mypyc-internal
  (linked into user extensions via SourceDep in mypyc/ir/deps.py); the
  librt.strings Python module shouldn't depend on them, matching how
  bytes_extra_ops, str_extra_ops, etc. are organized. librt.strings now
  picks up LibRTStrings_IsIdentifier via #include of the header.

Author: VaggelisD

mypyc/build.py

@@ -54,12 +54,7 @@ class ModDesc(NamedTuple):
 
 LIBRT_MODULES = [
     ModDesc("librt.internal", ["internal/librt_internal.c"], [], ["internal"]),
-    ModDesc(
-        "librt.strings",
-        ["strings/librt_strings.c", "codepoint_extra_ops.c"],
-        ["codepoint_extra_ops.h"],
-        ["strings"],
-    ),
+    ModDesc("librt.strings", ["strings/librt_strings.c"], [], ["strings"]),
     ModDesc(
         "librt.base64",
         [

mypyc/lib-rt/codepoint_extra_ops.c

@@ -1,20 +1,6 @@
+// All codepoint helper bodies live in codepoint_extra_ops.h as static
+// inline. This translation unit exists so the header is pulled into
+// mypyc-compiled extensions via SourceDep("codepoint_extra_ops.c") in
+// mypyc/ir/deps.py (which, in include_runtime_files mode, emits
+// `#include <codepoint_extra_ops.c>` into the generated __native.c).
 #include "codepoint_extra_ops.h"
-
-// Out-of-line bodies for codepoint helpers that are too large to inline.
-// The classification helpers and the ASCII fast paths for case conversion
-// stay inline in codepoint_extra_ops.h; this file holds the slow paths
-// that round-trip through PyUnicode_FromOrdinal and CPython's Unicode
-// machinery.
-
-bool LibRTStrings_IsIdentifier_slow(int32_t c) {
-    PyObject *s = PyUnicode_FromOrdinal((int)c);
-    if (s == NULL) {
-        // OOM. Swallow and return false to keep the function ERR_NEVER;
-        // callers expect a defined answer, not a propagated exception.
-        PyErr_Clear();
-        return false;
-    }
-    int r = PyUnicode_IsIdentifier(s);
-    Py_DECREF(s);
-    return r == 1;
-}

mypyc/lib-rt/codepoint_extra_ops.h

@@ -4,6 +4,7 @@
 #include <Python.h>
 #include <stdbool.h>
 #include <stdint.h>
+#include "CPy.h"
 
 // Codepoint helpers for librt.strings.
 // Inputs are signed int32_t for compatibility with mypyc's i32 type.
@@ -25,23 +26,26 @@ static inline bool LibRTStrings_IsAlpha(int32_t c) {
     return c >= 0 && Py_UNICODE_ISALPHA((Py_UCS4)c);
 }
 
-// Slow path for non-ASCII isidentifier; defined out-of-line in
-// codepoint_extra_ops.c because it allocates and calls into CPython.
-bool LibRTStrings_IsIdentifier_slow(int32_t c);
-
 // True if c could start a valid identifier (matches XID_Start
 // semantics, which is what str.isidentifier reports for a 1-character
 // string). The ASCII fast path covers `[A-Za-z_]` inline; non-ASCII
 // delegates to PyUnicode_IsIdentifier for correct PEP 3131 handling.
-// Returns false on OOM in the slow path (the function stays ERR_NEVER).
+// Aborts via CPyError_OutOfMemory on allocation failure, so this helper
+// stays ERR_NEVER.
 static inline bool LibRTStrings_IsIdentifier(int32_t c) {
     if (c < 0) return false;
     if (c < 128) {
         return (c >= 'a' && c <= 'z')
             || (c >= 'A' && c <= 'Z')
             || c == '_';
     }
-    return LibRTStrings_IsIdentifier_slow(c);
+    PyObject *s = PyUnicode_FromOrdinal((int)c);
+    if (s == NULL) {
+        CPyError_OutOfMemory();
+    }
+    int r = PyUnicode_IsIdentifier(s);
+    Py_DECREF(s);
+    return r == 1;
 }
 
 #endif  // MYPYC_CODEPOINT_EXTRA_OPS_H

mypyc/lib-rt/setup.py

@@ -103,7 +103,6 @@ def run(self) -> None:
                 "librt.strings",
                 [
                     "strings/librt_strings.c",
-                    "codepoint_extra_ops.c",
                     "init.c",
                     "int_ops.c",
                     "exc_ops.c",