[clang] [clang-format] Fix crash on numeric literals with an incomplete exponent (PR #206594)

[via cfe-commits]

llvmorg-github-actions[bot] wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-clang-format

Author: Gustas Janušonis (LeGusto)

<details>
<summary>Changes</summary>

NumericLiteralInfo computed the exponent/suffix search start from positions that can exceed the length of the string being searched (e.g. for `1e`, `1.0e`, `0xa.p`, `0x`, or `0x_`), so StringRef::find_insensitive and find_if_not read out of bounds -- an assertion failure on assertions-enabled builds and a silent out-of-bounds read otherwise. Clamp each search start to the length of the string it indexes.

Fixes #<!-- -->206593

---
Full diff: https://github.com/llvm/llvm-project/pull/206594.diff


2 Files Affected:

- (modified) clang/lib/Format/NumericLiteralInfo.cpp (+9-5) 
- (modified) clang/unittests/Format/NumericLiteralCaseTest.cpp (+11) 


``````````diff
diff --git a/clang/lib/Format/NumericLiteralInfo.cpp b/clang/lib/Format/NumericLiteralInfo.cpp
index 81e6dd5e58bbc..0bd27f50efa8c 100644
--- a/clang/lib/Format/NumericLiteralInfo.cpp
+++ b/clang/lib/Format/NumericLiteralInfo.cpp
@@ -16,6 +16,7 @@
 
 #include "NumericLiteralInfo.h"
 #include "llvm/ADT/StringExtras.h"
+#include <algorithm>
 
 namespace clang {
 namespace format {
@@ -46,11 +47,13 @@ NumericLiteralInfo::NumericLiteralInfo(StringRef Text, char Separator) {
 
   // e.g. 1.e2 or 0xFp2
   const auto Pos = DotPos != StringRef::npos ? DotPos + 1 : BaseLetterPos + 2;
+  // Trim C++ user-defined suffix as in `1_Pa`.
+  const auto TrimmedText =
+      Separator == '\'' ? Text.take_front(Text.find('_')) : Text;
 
-  ExponentLetterPos =
-      // Trim C++ user-defined suffix as in `1_Pa`.
-      (Separator == '\'' ? Text.take_front(Text.find('_')) : Text)
-          .find_insensitive(IsHex ? 'p' : 'e', Pos);
+  // Clamp searches due to possible incomplete literals
+  ExponentLetterPos = TrimmedText.find_insensitive(
+      IsHex ? 'p' : 'e', std::min(Pos, TrimmedText.size()));
 
   const bool HasExponent = ExponentLetterPos != StringRef::npos;
   SuffixPos = Text.find_if_not(
@@ -58,7 +61,8 @@ NumericLiteralInfo::NumericLiteralInfo(StringRef Text, char Separator) {
         return (HasExponent || !IsHex ? isDigit : isHexDigit)(C) ||
                C == Separator;
       },
-      HasExponent ? ExponentLetterPos + 2 : Pos); // e.g. 1e-2f
+      HasExponent ? std::min(ExponentLetterPos + 2, Text.size())
+                  : std::min(Pos, Text.size())); // e.g. 1e-2f
 }
 
 } // namespace format
diff --git a/clang/unittests/Format/NumericLiteralCaseTest.cpp b/clang/unittests/Format/NumericLiteralCaseTest.cpp
index ecd230d73f692..afb321d8d1b14 100644
--- a/clang/unittests/Format/NumericLiteralCaseTest.cpp
+++ b/clang/unittests/Format/NumericLiteralCaseTest.cpp
@@ -340,6 +340,17 @@ TEST_F(NumericLiteralCaseTest, UnderScoreSeparatorLanguages) {
   verifyFormat("o = 0o0_10_010;", "o = 0O0_10_010;", Style);
 }
 
+TEST_F(NumericLiteralCaseTest, IncompleteLiteralDoesNotCrash) {
+  auto Style = getLLVMStyle();
+  Style.NumericLiteralCase.HexDigit = FormatStyle::NLCS_Upper;
+
+  verifyFormat("i = 1e;", Style);
+  verifyFormat("i = 1.0e;", Style);
+  verifyFormat("i = 0x1p;", Style);
+  verifyFormat("i = 0x;", Style);
+  verifyFormat("i = 0x_;", Style);
+}
+
 } // namespace
 } // namespace test
 } // namespace format

``````````

</details>


https://github.com/llvm/llvm-project/pull/206594