Date: Sun, 28 Aug 2016 04:15:01 +0300 From: Konstantin Belousov <kostikbel@gmail.com> To: Andrey Chernov <ache@freebsd.org> Cc: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: Re: svn commit: r304928 - in head/lib/libc: amd64/sys i386/sys sys Message-ID: <20160828011501.GH83214@kib.kiev.ua> In-Reply-To: <59ac1812-7c77-b677-51c4-dcadc6b2be7f@freebsd.org> References: <201608272303.u7RN3N0D078505@repo.freebsd.org> <9bcf10db-de3f-33ce-e418-03ce3283ac90@freebsd.org> <20160828005637.GG83214@kib.kiev.ua> <59ac1812-7c77-b677-51c4-dcadc6b2be7f@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Aug 28, 2016 at 03:50:04AM +0300, Andrey Chernov wrote: > On 28.08.2016 3:38, Andrey Chernov wrote: > > On 28.08.2016 2:03, Konstantin Belousov wrote: > >> Since ptrace(2) syscall can return -1 for non-error situations, libc > >> wrappers set errno to 0 before performing the syscall, as the service > >> to the caller. > > > > Both C99 and POSIX directly prohibits any standard function to set errno > > to 0. ptrace() should either choose other errno to indicate non-error > > situation or change return -1 to something else. > > > ...and don't touch errno. > > POSIX: "No function in this volume of POSIX.1-2008 shall set errno to zero." I am quite curious where ptrace(2) is defined by POSIX. > > > On both i386 and amd64, the errno symbol was directly > > referenced, which only works correctly in single-threaded process. > > POSIX: "For each thread of a process, the value of errno shall not be > affected by function calls or assignments to errno by other threads." And ? What should the citation add new to the substance of the code change ? On Sun, Aug 28, 2016 at 04:04:00AM +0300, Andrey Chernov wrote: > On 28.08.2016 3:56, Konstantin Belousov wrote: > > On Sun, Aug 28, 2016 at 03:38:10AM +0300, Andrey Chernov wrote: > >> On 28.08.2016 2:03, Konstantin Belousov wrote: > >>> Since ptrace(2) syscall can return -1 for non-error situations, libc > >>> wrappers set errno to 0 before performing the syscall, as the service > >>> to the caller. > >> > >> Both C99 and POSIX directly prohibits any standard function to set errno > >> to 0. ptrace() should either choose other errno to indicate non-error > >> situation or change return -1 to something else. > >> > > ptrace(2) is not a standard function. > > And, we cannot break ABI for the syscall. > > > > C99 statement sounds stricter: > "The value of errno is zero at program startup, but is never set to zero > by any library function. 176)" > And syscall is not different from library function from C99 point of view. Point me to a single line in C99 which mentions ptrace(). Do you understand what did the commit changed, and what it did not ? Setting errno to zero before the syscall was the existing behaviour before the change, and I did not modified anything there. But previous wrapper set errno to zero in main thread even if called from some other thread, which was the bug fixed.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20160828011501.GH83214>