- mips POLLWRNORM and POLLWRBAND macros had wrong values
- x32 pthread synchronization object type definitions were wrong
- powerpc minimum signal stack size was insufficient
+
+
+
+1.1.8 release notes
+
+bugs fixed:
+- stack-based buffer overflow in inet_pton (CVE-2015-1817)
+- regcomp crash/mem-corruption with illegal bytes after backslash
+- regcomp wrongly allowed backrefs in ER
+- regcomp miscompiled character class brace-repetitions
+- regcomp wrongly processed \0 as an unmatchable backref
+- new FLT_ROUNDS definition failed to work in C++ code
+
+arch-specific bugs fixed:
+- aarch64 was missing max_align_t definition
+
+
+
+1.1.9 release notes
+
+new features:
+- ability to protect libc code itself with stack protector
+- sigsetjmp now restores signal mask after restoring context, not before
+- thread-local dlerror status/messages
+- dlerror messages are no longer truncated
+- diagnostics for constraint violations with ctype.h macros
+
+optimizations:
+- reduce cost of PIC on archs where PLT calls need a fixed GOT register
+- spin locks no longer constantly invalidate cache lines while spinning
+- code size reduction in static-linked TLS init
+
+bugs fixed:
+- failure to process robust mutexes on detached-thread exit
+- possible memory corruption due to robust mutex list on detached-thread exit
+- crash on memory exhaustion in getgr* internals
+- misaligned memory accesses in static binaries with low-alignment TLS blocks
+- multiple cases of wrongful path search continuation after transient failure
+- small memory leak on failure of dlopen with RPATH $ORIGIN
+- several small math bugs related to exception flags with non-finite args
+- mmap leak in sem_open failure path for link call
+- duplocale clobbered new locale struct with memcpy of old
+- futimes crashed with null timeval argument
+
+arch-specific bugs fixed:
+- stack protector spuriously aborted after forking on x32
+- stack protector spuriously aborted with flockfile on powerpc
+- theoretically-possible clobbering of syscall return value on mips
+- random thread-pointer setup failure on sh (uninitialized return value)
+- possible crash in dlsym on sh due to incorrectly-computed branch target
+- broken fesetenv(FE_DFL_ENV) on mips
+- dynamic linker name for sh ignored fpu/nofpu and endianness
+- various minor aarch64 bugs
+- dangling pointers in x32 syscall timespec fixup code
+
+
+
+1.1.10 release notes
+
+new features:
+- fail-safe (allocation-free) C locale for newlocale to return
+- all locale categories track requested locale name
+- rcrt1.o start file for static PIE
+
+optimizations:
+- inline atomics for sh4a
+- removed heavy atomics from locale-related code paths
+- removed global data accesses from CURRENT_LOCALE macro & callers
+- dynamic linker stage 1 size reduction
+
+compatibility:
+- better configure detection of unsupported compiler options
+- support for more relocation types in libc.so, not currently used
+- iconv_open accepts "" and "CHAR" as aliases for native (UTF-8)
+- additional LFS64 macros in sys/resource.h
+
+regressions fixed:
+- dynamic linker crash on NONE-type relocations (only mips affected)
+- inability to build as thumb2 on arm
+- failure to run under qemu-i386 user-level emulation
+- inability to access globals from libc on powerpc
+- PIE link errors in Scrt1.o under unusual usage on some archs
+
+other bugs fixed:
+- failure of ungetc/ungetwc to work on FILE streams in EOF state
+- possible null pointer dereference in gettext
+- possible initial stack misalignment on mips with PIE
projects / musl / blobdiff