[libc] Fix typo in long double negative block

[michaelrj-google]

The long double version of float to string's get_negative_block had a
bug in table mode. In table mode, one of the tables is named
"MIN_BLOCK_2" and it stores the number of blocks that are all zeroes
before the digits start for a given index. The check for long doubles
was incorrectly "block_index <= MIN_BLOCK_2[idx]" when it should be
"block_index < MIN_BLOCK_2[idx]" (without the equal sign). This bug
caused an off-by-one error for some long double values. This patch fixes
the bug and adds tests to ensure it doesn't regress.

[llvmbot]

@llvm/pr-subscribers-libc

Changes

The long double version of float to string's get_negative_block had a
bug in table mode. In table mode, one of the tables is named
"MIN_BLOCK_2" and it stores the number of blocks that are all zeroes
before the digits start for a given index. The check for long doubles
was incorrectly "block_index <= MIN_BLOCK_2[idx]" when it should be
"block_index < MIN_BLOCK_2[idx]" (without the equal sign). This bug
caused an off-by-one error for some long double values. This patch fixes
the bug and adds tests to ensure it doesn't regress.

---

Full diff: https://github.com/llvm/llvm-project/pull/68243.diff

2 Files Affected:

• (modified) libc/src/__support/float_to_string.h (+1-1)
• (modified) libc/test/src/stdio/sprintf_test.cpp (+11)

diff --git a/libc/src/__support/float_to_string.h b/libc/src/__support/float_to_string.h
index d3fcec7a652faba..eb06cd9c08af287 100644
--- a/libc/src/__support/float_to_string.h
+++ b/libc/src/__support/float_to_string.h
@@ -708,7 +708,7 @@ FloatToString<long double>::get_negative_block(int block_index) {
     const int32_t SHIFT_CONST = TABLE_SHIFT_CONST;
 
     // if the requested block is zero
-    if (block_index <= MIN_BLOCK_2[idx]) {
+    if (block_index < MIN_BLOCK_2[idx]) {
       return 0;
     }
     const uint32_t p = POW10_OFFSET_2[idx] + block_index - MIN_BLOCK_2[idx];
diff --git a/libc/test/src/stdio/sprintf_test.cpp b/libc/test/src/stdio/sprintf_test.cpp
index 8b9c919bed203d4..63de6f94dc1ab5e 100644
--- a/libc/test/src/stdio/sprintf_test.cpp
+++ b/libc/test/src/stdio/sprintf_test.cpp
@@ -2757,6 +2757,17 @@ TEST_F(LlvmLibcSPrintfTest, FloatAutoConv) {
   written = LIBC_NAMESPACE::sprintf(buff, "%.10g", 0x1.0p-1074);
   ASSERT_STREQ_LEN(written, buff, "4.940656458e-324");
 
+  written = LIBC_NAMESPACE::sprintf(buff, "%g", 0xa.aaaaaaaaaaaaaabp-7);
+  ASSERT_STREQ_LEN(written, buff, "0.0833333");
+
+  written = LIBC_NAMESPACE::sprintf(buff, "%Lg", 0xa.aaaaaaaaaaaaaabp-7L);
+  ASSERT_STREQ_LEN(written, buff, "0.0833333");
+
+  written = LIBC_NAMESPACE::sprintf(buff, "%.60Lg", 0xa.aaaaaaaaaaaaaabp-7L);
+  ASSERT_STREQ_LEN(
+      written, buff,
+      "0.0833333333333333333355920878593448009041821933351457118988037");
+
   // Long double precision tests.
   // These are currently commented out because they require long double support
   // that isn't ready yet.

[lntue]

Should these tests be guarded in the ifdefs below?