From 8c1c57a64b664340b0ede1c1f0eff8212b475ef5 Mon Sep 17 00:00:00 2001 From: Rich Felker Date: Wed, 18 Mar 2015 20:38:02 -0400 Subject: release 1.1.7 --- WHATSNEW | 43 +++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 43 insertions(+) (limited to 'WHATSNEW') diff --git a/WHATSNEW b/WHATSNEW index 6fa87652..812b3ac0 100644 --- a/WHATSNEW +++ b/WHATSNEW @@ -1410,3 +1410,46 @@ bugs fixed: arch-specific bugs fixed: - on or1k, some syscalls with 64-bit arguments were broken (misaligned) - usage of sahf instruction on x86_64 crashed on some early cpu models + + + +1.1.7 release notes + +new features: +- alternate passwd/group backend support via nscd protocol +- masked cancellation mode extension (experimental) +- aio cancellation +- aarch64 port (experimental) + +performance: +- significant memset asm optimizations on i386 and x86_64 + +compatibility: +- suppress EINTR in semaphores for old kernels where futex restart is broken +- always set optarg in getopt_long +- support SOCK_RAW socket type in getaddrinfo +- report success instead of EINPROGRESS when close is interrupted + +bugs fixed: +- multithreaded set*id() was not async-signal safe, had various race bugs +- getspnam_r returned results for partial username matches +- wordexp bad character checker mis-counted parentheses +- close on fd with pending aio could lead to file corruption +- old aio implementation had numerous conformance bugs +- malloc init code could deadlock due to race condition +- pthread_exit did not disable cancellation +- pthread_cond_wait could wrongly consume signal on cancellation +- execvp wrongly stopped path search on EACCESS +- fsync, fdatasync, and msync were not honored as cancellation points +- fchmodat was subject to fd leak race (missing O_CLOEXEC) +- fchmodat failed to report EOPNOTSUPP in race path +- passwd/group lookup functions had various minor error-reporting bugs +- isatty had false-positives/device-state-corruption for OSS sound devices +- configure script failed to detect gcc with translated messages +- FLT_ROUNDS macro failed to reflect rounding mode changes in fenv + +arch-specific bugs fixed: +- mips fesetenv did not handle FE_DFL_ENV +- mips POLLWRNORM and POLLWRBAND macros had wrong values +- x32 pthread synchronization object type definitions were wrong +- powerpc minimum signal stack size was insufficient -- cgit v1.2.1