Date: Tue, 5 Jul 2016 15:55:12 -0500 From: Pedro Giffuni <pfg@FreeBSD.org> To: cem@freebsd.org, "Ngie Cooper (yaneurabeya)" <yaneurabeya@gmail.com> Cc: Dimitry Andric <dim@freebsd.org>, Roman Divacky <rdivacky@freebsd.org>, FreeBSD current <freebsd-current@freebsd.org> Subject: Re: clang 3.3/3.4 fails to build items that use stdlib.h because of __alloc_size attribute assigned to posix_memalign Message-ID: <647ae486-0639-715e-4fc2-28590295f816@FreeBSD.org> In-Reply-To: <3a288119-81f5-9199-b9d4-0f4649a7b8eb@FreeBSD.org> References: <838A69F7-343D-4398-928B-E54FB966B574@gmail.com> <CAG6CVpXHK1J7vvR1LgFLs_fGyoZTTVw1Jput5xswKRSeYHEquw@mail.gmail.com> <3a288119-81f5-9199-b9d4-0f4649a7b8eb@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Answering to myself .... On 07/05/16 15:24, Pedro Giffuni wrote: > Hmm ... > > On 07/05/16 15:14, Conrad Meyer wrote: >> Whoops, missed reply-all the first time. >> >> It seems pretty clear that alloc_size (return value is a memory >> allocation of size from parameter N) does not apply to posix_memalign, >> because posix_memalign's allocation is stored via a pointer argument >> rather than return value. >> >> https://gcc.gnu.org/onlinedocs/gcc/Common-Function-Attributes.html#Common-Function-Attributes >> >> >> IMO the attribute should be removed from posix_memalign in stdlib.h >> >> Best, >> Conrad >> > > - I am wondering why it hasn't affected -current if it is wrong. Because newer clang removed alloc_size altogether. > - I am also wondering if we shouldn't just get rid of the attribute. > Nah, it is basically meant for FORTIFY_SOURCE, which is not being worked on anymore but maybe handy some day. > I will answer to myself those doubts and fix the issue. Thanks! > So, yes, we should remove it from posix_memalign(). I'll ask re@. Pedro.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?647ae486-0639-715e-4fc2-28590295f816>