[clang] [analyzer] Don't fold wide string literals in CStringChecker (PR #213281)

via cfe-commits cfe-commits at lists.llvm.org
Fri Jul 31 07:25:29 PDT 2026


llvmorg-github-actions[bot] wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-clang

Author: Balázs Benics (steakhal)

<details>
<summary>Changes</summary>

`getStringRefAtRegion()` read literal contents via `StringLiteral::getBytes()`, which returns the literal's raw trailing storage. For wide literals (`u""`, `U""`, `L""`) that storage holds the code units in *host* byte order: `LiteralSupport` writes them through `reinterpret_cast<UTF16*>/<UTF32*>`, and `StringLiteral::getCodeUnit()` reads them back the same way. Interpreting those bytes as a target byte string made the strchr-family constant folding depend on the endianness of the machine running the analyzer.

Bail out on `getCharByteWidth() != 1`, matching the contract the rest of the checker already relies on -- `evalStrcmpCommon()` uses `getString()`, which asserts it. Wide literals now fall back to the symbolic-offset path and keep both branches.

Fixes up #<!-- -->212124
Inspired from:
https://github.com/llvm/llvm-project/pull/212124#issuecomment-5143598098

Supersedes #<!-- -->213279

---
Full diff: https://github.com/llvm/llvm-project/pull/213281.diff


2 Files Affected:

- (modified) clang/lib/StaticAnalyzer/Checkers/CStringChecker.cpp (+5) 
- (modified) clang/test/Analysis/string-search-modeling.c (+11-30) 


``````````diff
diff --git a/clang/lib/StaticAnalyzer/Checkers/CStringChecker.cpp b/clang/lib/StaticAnalyzer/Checkers/CStringChecker.cpp
index b47fca0e40af4..9fd8d4880c59f 100644
--- a/clang/lib/StaticAnalyzer/Checkers/CStringChecker.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/CStringChecker.cpp
@@ -1075,6 +1075,11 @@ CStringChecker::getStringRefAtRegion(const MemRegion *R) {
   const StringLiteral *Lit = getStringLiteralFromRegion(Base);
   if (!Lit)
     return std::nullopt;
+  // getBytes() exposes the literal's raw storage, which for wide literals holds
+  // the code units in host byte order (see StringLiteral::getCodeUnit()).
+  // Only narrow literals can be interpreted as a target byte string.
+  if (Lit->getCharByteWidth() != 1)
+    return std::nullopt;
   StringRef S = Lit->getBytes();
   if (Offset > S.size())
     return std::nullopt;
diff --git a/clang/test/Analysis/string-search-modeling.c b/clang/test/Analysis/string-search-modeling.c
index f18ffde3d198d..a43a836a4bf44 100644
--- a/clang/test/Analysis/string-search-modeling.c
+++ b/clang/test/Analysis/string-search-modeling.c
@@ -1,5 +1,8 @@
 // The u"" and U"" string literals below need C11 or later. Pin the standard
 // because targets such as PS4 default to gnu99.
+//
+// Deliberately no -triple here: the expectations below must hold for every
+// target and host, including big-endian ones. See test_strchr_wide_string_global.
 // RUN: %clang_analyze_cc1 -std=c17 -verify %s \
 // RUN:   -analyzer-checker=core,unix \
 // RUN:   -analyzer-checker=debug.ExprInspection \
@@ -522,53 +525,31 @@ void test_strpbrk_accept_embedded_null(void) {
   clang_analyzer_eval(strpbrk("xb", "a\0b") == 0); // expected-warning {{TRUE}}
 }
 
-// --- Wide string cast to char*: resolved via raw bytes ---
+// --- Wide string cast to char*: not folded ---
+// A wide literal's AST storage holds its code units in host byte order, so its
+// raw bytes must not be interpreted as a target byte string. The checker bails
+// out on these and stays conservative, keeping both branches.
 const __CHAR16_TYPE__ wide_str_global[] = u"abc";
 void test_strchr_wide_string_global(void) {
-#if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
-  // LE: bytes are 'a',0,'b',0,... — CStr is "a", strchr finds 'a' at offset 0.
-  clang_analyzer_eval(strchr((const char *)wide_str_global, 'a') == (const char *)wide_str_global); // expected-warning {{TRUE}}
-#else
-  // BE: bytes are 0,'a',0,'b',... — first byte is null, CStr is empty.
-  clang_analyzer_eval(strchr((const char *)wide_str_global, 'a') == 0); // expected-warning {{TRUE}}
-#endif
+  clang_analyzer_eval(strchr((const char *)wide_str_global, 'a') == 0); // expected-warning {{TRUE}} expected-warning {{FALSE}}
 }
 
 void test_strchr_wide_string_local(void) {
   const __CHAR16_TYPE__ w[] = u"abc";
-#if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
-  // LE: same as global — finds 'a' at offset 0.
-  clang_analyzer_eval(strchr((const char *)w, 'a') == (const char *)w); // expected-warning {{TRUE}}
-#else
-  // BE: first byte is null, CStr is empty.
-  clang_analyzer_eval(strchr((const char *)w, 'a') == 0); // expected-warning {{TRUE}}
-#endif
+  clang_analyzer_eval(strchr((const char *)w, 'a') == 0); // expected-warning {{TRUE}} expected-warning {{FALSE}}
 }
 
 // --- Wide string with 4-byte characters (UTF-32) ---
 const __CHAR32_TYPE__ wide32_global[] = U"abc";
 void test_strchr_wide32_string(void) {
-#if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
-  // LE: U"abc" bytes are 'a',0,0,0,'b',0,0,0,...  CStr is "a".
-  clang_analyzer_eval(strchr((const char *)wide32_global, 'a') == (const char *)wide32_global); // expected-warning {{TRUE}}
-#else
-  // BE: bytes are 0,0,0,'a',... — first byte is null, CStr is empty.
-  clang_analyzer_eval(strchr((const char *)wide32_global, 'a') == 0); // expected-warning {{TRUE}}
-#endif
+  clang_analyzer_eval(strchr((const char *)wide32_global, 'a') == 0); // expected-warning {{TRUE}} expected-warning {{FALSE}}
 }
 
 // --- Wide string as needle argument ---
 const __CHAR16_TYPE__ wide_needle[] = u"lo";
 void test_strstr_wide_needle(void) {
   const char *s = "hello";
-#if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
-  // LE: u"lo" bytes are 'l',0,'o',0,0,0 — getCStr gives "l".
-  // strstr("hello", "l") finds 'l' at offset 2.
-  clang_analyzer_eval(strstr(s, (const char *)wide_needle) == s + 2); // expected-warning {{TRUE}}
-#else
-  // BE: bytes are 0,'l',0,'o',... — CStr is empty, strstr returns haystack.
-  clang_analyzer_eval(strstr(s, (const char *)wide_needle) == s); // expected-warning {{TRUE}}
-#endif
+  clang_analyzer_eval(strstr(s, (const char *)wide_needle) == 0); // expected-warning {{TRUE}} expected-warning {{FALSE}}
 }
 
 // --- Struct cast to char* ---

``````````

</details>


https://github.com/llvm/llvm-project/pull/213281


More information about the cfe-commits mailing list