[libc] Disable -NaN test on float128 systems

[michaelrj-google]

Some float128 systems (specifically the ones used for aarch64 buildbots)
don't respect signs for long double NaNs. This patch disables the printf
test that was failing due to this.

[llvmbot]

@llvm/pr-subscribers-libc

Author: None (michaelrj-google)

Changes

Some float128 systems (specifically the ones used for aarch64 buildbots)
don't respect signs for long double NaNs. This patch disables the printf
test that was failing due to this.

---

Full diff: https://github.com/llvm/llvm-project/pull/70146.diff

1 Files Affected:

• (modified) libc/test/src/stdio/sprintf_test.cpp (+4)

diff --git a/libc/test/src/stdio/sprintf_test.cpp b/libc/test/src/stdio/sprintf_test.cpp
index 5aa346c414b1fe1..a68ee97d6d6edc4 100644
--- a/libc/test/src/stdio/sprintf_test.cpp
+++ b/libc/test/src/stdio/sprintf_test.cpp
@@ -1008,8 +1008,12 @@ TEST_F(LlvmLibcSPrintfTest, FloatDecimalConv) {
   written = LIBC_NAMESPACE::sprintf(buff, "%Lf", ld_nan);
   ASSERT_STREQ_LEN(written, buff, "nan");
 
+// Some float128 systems (specifically the ones used for aarch64 buildbots)
+// don't respect signs for long double NaNs.
+#if defined(SPECIAL_X86_LONG_DOUBLE) || defined(LONG_DOUBLE_IS_DOUBLE)
   written = LIBC_NAMESPACE::sprintf(buff, "%LF", -ld_nan);
   ASSERT_STREQ_LEN(written, buff, "-NAN");
+#endif
 
   // Length Modifier Tests.
 

[michaelrj-google]

Landed without review as a trivial change to fix the buildbots.

[lntue]

Thanks for taking care of this! Probably we should add comments about aarch64's behavior in the test, and add a test to check only NaN containment in the #else clause in the followup.