[NFC] [analyzer] Add ArrayBound tests to document casting bug

[NagyDonat]

Add a few security.ArrayBound testcases that document the false positives caused the fact that the analyzer doesn't model a cast from signed char to unsigned char.

[llvmbot]

@llvm/pr-subscribers-clang

Author: Donát Nagy (NagyDonat)

Changes

Add a few security.ArrayBound testcases that document the false positives caused the fact that the analyzer doesn't model a cast from signed char to unsigned char.

---

Full diff: https://github.com/llvm/llvm-project/pull/127062.diff

1 Files Affected:

• (modified) clang/test/Analysis/out-of-bounds.c (+31)

diff --git a/clang/test/Analysis/out-of-bounds.c b/clang/test/Analysis/out-of-bounds.c
index 923797200d0b4..177c03500a0c3 100644
--- a/clang/test/Analysis/out-of-bounds.c
+++ b/clang/test/Analysis/out-of-bounds.c
@@ -194,3 +194,34 @@ char test_comparison_with_extent_symbol(struct incomplete *p) {
   return ((char *)p)[-1]; // no-warning
 }
 
+int table[256], small_table[128];
+int test_cast_to_unsigned(signed char x) {
+  unsigned char y = x;
+  if (x >= 0)
+    return x;
+  // FIXME: Here the analyzer ignores the signed -> unsigned cast, and manages to
+  // load a negative value from an unsigned variable. This causes an underflow
+  // report, which is an ugly false positive.
+  // The underlying issue is tracked by Github ticket #39492.
+  return table[y]; // expected-warning {{Out of bound access to memory preceding}}
+}
+
+int test_cast_to_unsigned_overflow(signed char x) {
+  unsigned char y = x;
+  if (x >= 0)
+    return x;
+  // A variant of 'test_cast_to_unsigned' where the correct behavior would be
+  // an overflow report (because the negative values are cast to `unsigned
+  // char` values that are too large).
+  // FIXME: See comment in 'test_cast_to_unsigned'.
+  return small_table[y]; // expected-warning {{Out of bound access to memory preceding}}
+}
+
+int test_negative_offset_with_unsigned_idx(void) {
+  // An example where the subscript operator uses an unsigned index, but the
+  // underflow report is still justified. (We should try to keep this if we
+  // silence false positives like the one in 'test_cast_to_unsigned'.)
+  int *p = table - 10;
+  unsigned idx = 2u;
+  return p[idx]; // expected-warning {{Out of bound access to memory preceding}}
+}

[steakhal]

Makes sense.

[steakhal]

Can we also dump the value range of this y sym?
That would probably highlight the "wrong assumption"

Maybe in the rest of the examples too. Wdyt?

[NagyDonat]

Yep, that seems to be useful, I added the value dumps to the two testcases where they are relevant. (The third testcase is not buggy, I just created it to highlight potential drawbacks of naive solutions for silencing the bug.)

[steakhal]

Looks great. Thanks.