https://bugs.kde.org/show_bug.cgi?id=466104
--- Comment #2 from Paul Floyd <pjfl...@wanadoo.fr> --- More details. glibc memalign is here https://elixir.bootlin.com/glibc/glibc-2.37/source/malloc/malloc.c#L3504 if (alignment <= MALLOC_ALIGNMENT) return __libc_malloc (bytes); /* Otherwise, ensure that it is at least a minimum chunk size */ if (alignment < MINSIZE) alignment = MINSIZE; /* If the alignment is greater than SIZE_MAX / 2 + 1 it cannot be a power of 2 and will cause overflow in the check below. */ if (alignment > SIZE_MAX / 2 + 1) { __set_errno (EINVAL); return 0; } What are those macros? #define MALLOC_ALIGNMENT (2 * SIZE_SZ < __alignof__ (long double) \ ? __alignof__ (long double) : 2 * SIZE_SZ) SIZE_SZ is the size of size_t, 8 on 64bit and 4 on 32bit. alignof long double is 4 on x86 16 on amd64 16 on aarch64 8/16 for MIPS32/64 and 16 for PPC. I don't think that the condition is true on any platform we support so this is 8 on 32bit and 16 on 64bit. MINSIZE, another macro rabbit hole #define MINSIZE \ (unsigned long)(((MIN_CHUNK_SIZE+MALLOC_ALIGN_MASK) & ~MALLOC_ALIGN_MASK)) MIN_CHUNK_SIZE is 16/32 on 32/64 bit systems. MALLOC_ALIGN_MASK is MALLOC_ALIGNMENT - 1, so 7 or 15 I make that 16 for 32bit systems and 32 for 64bit systems. SIZE_MAX is in stdint.h. I can add a VKI_SIZE_MAX. -- You are receiving this mail because: You are watching all bug changes.
