[libc++] Remove potential 0-sized array in __compressed_pair_padding

[serge-sans-paille]

No description provided.

[llvmbot]

@llvm/pr-subscribers-clang

@llvm/pr-subscribers-libcxx

Author: None (serge-sans-paille)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/109028.diff

1 Files Affected:

• (modified) libcxx/include/__memory/compressed_pair.h (+6-4)

diff --git a/libcxx/include/__memory/compressed_pair.h b/libcxx/include/__memory/compressed_pair.h
index 629e3ad8848ffa..cc56912f5e34c5 100644
--- a/libcxx/include/__memory/compressed_pair.h
+++ b/libcxx/include/__memory/compressed_pair.h
@@ -52,13 +52,15 @@ _LIBCPP_BEGIN_NAMESPACE_STD
 
 #ifndef _LIBCPP_ABI_NO_COMPRESSED_PAIR_PADDING
 
-template <class _ToPad>
+template <class _ToPad,
+          bool _Empty = ((is_empty<_ToPad>::value && !__libcpp_is_final<_ToPad>::value) || is_reference<_ToPad>::value)>
 class __compressed_pair_padding {
-  char __padding_[((is_empty<_ToPad>::value && !__libcpp_is_final<_ToPad>::value) || is_reference<_ToPad>::value)
-                      ? 0
-                      : sizeof(_ToPad) - __datasizeof_v<_ToPad>];
+  char __padding_[sizeof(_ToPad) - __datasizeof_v<_ToPad>];
 };
 
+template <class _ToPad>
+class __compressed_pair_padding<_ToPad, true> {};
+
 #  define _LIBCPP_COMPRESSED_PAIR(T1, Initializer1, T2, Initializer2)                                                  \
     _LIBCPP_NO_UNIQUE_ADDRESS __attribute__((__aligned__(_LIBCPP_ALIGNOF(T2)))) T1 Initializer1;                       \
     _LIBCPP_NO_UNIQUE_ADDRESS ::std::__compressed_pair_padding<T1> _LIBCPP_CONCAT3(__padding1_, __LINE__, _);          \

[serge-sans-paille]

In the spirit of #105865 - I don't think we can use a similar implementation as what we did there, but that's the same idea / motivation

[ldionne]

Can't we reuse the same __padding struct we now have in <string>?

[serge-sans-paille]

indeed

[serge-sans-paille]

It's not clear to me if the issue here come from the CI system or from my patch :-/

[shafik]

This needs a more detailed summary, e.g. what is the motivation for this change?

[ldionne]

Please rebase onto main to re-trigger CI. The CI instability should be resolved now.

[github-actions]

✅ With the latest revision this PR passed the C/C++ code formatter.

[serge-sans-paille]

@shafik : I'm trying to have libcxx compile with -Wzero-length-array. With these two commits, we're done.

@ldionne : I undid the reuse of __padding. The motivation is two folds: 1st it would create an extra header dependency (moving __padding definition to another header so that it can be included both in string and compressed_pair.h) and second compressed_pair is now scarcely used, so it's probably ok to keep related stuff in the same header (?).