[PATCH] PR libstdc++/90046 fix build failure on epiphany-elf

[PATCH] PR libstdc++/90046 fix build failure on epiphany-elf

[Jonathan Wakely]

[-- Attachment #1: Type: text/plain, Size: 918 bytes --]

The epiphany-elf target aligns structs to 8 bytes, which causes the
static_assert(alignof(_Chunk) == 1) to fail.

Instead of requiring _Chunks to be positionable at any alignment, ensure
new buffers are aligned to alignof(_Chunk). Because the buffer size is a
power of two, we know that both the buffer size and sizeof(_Chunk) are
multiples of alignof(_Chunk). So is p is aligned to alignof(_Chunk) then
so is (p + size - sizeof(_Chunk)). So just ensure the new buffer is
aligned to at least alignof(_Chunk), which should already be true
because the caller requests at least alignof(max_align_t).

	PR libstdc++/90046
	* src/c++17/memory_resource.cc
	(monotonic_buffer_resource::_Chunk::allocate): Increase alignment if
	needed to allow placing a _Chunk at the end of the buffer.
	(monotonic_buffer_resource::_M_new_buffer): Remove static_assert.

Tested x86_64-linux, bootstrapped epiphany-elf, committed to trunk./


[-- Attachment #2: patch.txt --]
[-- Type: text/plain, Size: 2630 bytes --]

commit 8af70056b079035bfb5a6a3a9df4b10194b025cf
Author: Jonathan Wakely <jwakely@redhat.com>
Date:   Thu Apr 11 20:32:40 2019 +0100

    PR libstdc++/90046 fix build failure on epiphany-elf
    
    The epiphany-elf target aligns structs to 8 bytes, which causes the
    static_assert(alignof(_Chunk) == 1) to fail.
    
    Instead of requiring _Chunks to be positionable at any alignment, ensure
    new buffers are aligned to alignof(_Chunk). Because the buffer size is a
    power of two, we know that both the buffer size and sizeof(_Chunk) are
    multiples of alignof(_Chunk). So is p is aligned to alignof(_Chunk) then
    so is (p + size - sizeof(_Chunk)). So just ensure the new buffer is
    aligned to at least alignof(_Chunk), which should already be true
    because the caller requests at least alignof(max_align_t).
    
            PR libstdc++/90046
            * src/c++17/memory_resource.cc
            (monotonic_buffer_resource::_Chunk::allocate): Increase alignment if
            needed to allow placing a _Chunk at the end of the buffer.
            (monotonic_buffer_resource::_M_new_buffer): Remove static_assert.

diff --git a/libstdc++-v3/src/c++17/memory_resource.cc b/libstdc++-v3/src/c++17/memory_resource.cc
index cd11bf5875c..b6698011f5c 100644
--- a/libstdc++-v3/src/c++17/memory_resource.cc
+++ b/libstdc++-v3/src/c++17/memory_resource.cc
@@ -182,7 +182,21 @@ namespace pmr
 	     _Chunk*& __head)
     {
       __size = std::__ceil2(__size + sizeof(_Chunk));
+
+      if constexpr (alignof(_Chunk) > 1)
+	{
+	  // PR libstdc++/90046
+	  // For targets like epiphany-elf where alignof(_Chunk) != 1
+	  // ensure that the last sizeof(_Chunk) bytes in the buffer
+	  // are suitably-aligned for a _Chunk.
+	  // This should be unnecessary, because the caller already
+	  // passes in max(__align, alignof(max_align_t)).
+	  if (__align < alignof(_Chunk))
+	    __align = alignof(_Chunk);
+	}
+
       void* __p = __r->allocate(__size, __align);
+
       // Add a chunk defined by (__p, __size, __align) to linked list __head.
       void* const __back = (char*)__p + __size - sizeof(_Chunk);
       __head = ::new(__back) _Chunk(__size, __align, __head);
@@ -231,9 +245,6 @@ namespace pmr
   void
   monotonic_buffer_resource::_M_new_buffer(size_t bytes, size_t alignment)
   {
-    // Need to check this somewhere, so put it here:
-    static_assert(alignof(monotonic_buffer_resource::_Chunk) == 1);
-
     const size_t n = std::max(bytes, _M_next_bufsiz);
     const size_t m = std::max(alignment, alignof(std::max_align_t));
     auto [p, size] = _Chunk::allocate(_M_upstream, n, m, _M_head);

Re: [PATCH] PR libstdc++/90046 fix build failure on epiphany-elf

[Jonathan Wakely]

On 11/04/19 20:58 +0100, Jonathan Wakely wrote:
>The epiphany-elf target aligns structs to 8 bytes, which causes the
>static_assert(alignof(_Chunk) == 1) to fail.
>
>Instead of requiring _Chunks to be positionable at any alignment, ensure
>new buffers are aligned to alignof(_Chunk). Because the buffer size is a
>power of two, we know that both the buffer size and sizeof(_Chunk) are
>multiples of alignof(_Chunk). So is p is aligned to alignof(_Chunk) then

Oops, that was meant to say "So if p is aligned to ..."