Re: draft release notes for 0.9.15

[Szabolcs Nagy <nsz@port70.net>]

* Rich Felker <dalias@aerifal.cx> [2013-12-03 20:33:20 -0500]:
> See draft below. Comments welcome, especially on what's the most
> important to go in the short release blurb since there's so much..

my list would be:

new features:
v4 and v6 nameserver in resolv.conf
multicast structures in netinet/in.h
shadow password api
libc.so can print musl version info

bug fixes:
mbsrtowcs buffer overflow
group file handling
execle environ passing
setenv crash
timezone
ip address parsing
faccessat
fnmatch
fd leaks

> DRAFT 0.9.15 release notes
> 
> new features:
> - support for mixing IPv4 and v6 nameserver addresses in resolv.conf
> - RFC 3678 multicast structures/macros in netinet/in.h
> - putspent and fgetspent functions (shadow password API)
> - timef function (obsolete, removed in POSIX 2008)
> - semtimedop syscall (Linux-specific sysvipc extension)
> - drem and finite functions (obsolete BSD functions)
> - getloadavg function (non-standard)
> - libc.so now stores and prints its version information
> - expose constants for new Linux features including O_TMPFILE
> 
> bugs fixed:

- fnmatch out-of-bounds access and spurious failures with FNM_PATHNAME and escaped / pattern

> - buffer overflow in mbsrtowcs
> - clobbering of gr_name in getgrnam_r and getgrgid_r
> - execle ignoring the environment argument
> - setenv crash on malloc failure
> - failure of malloc to set errno when failing to extend heap
> - incorrect errno value from getcwd with zero size
> - spurious failure in faccessat with AT_EACCESS flag with suid/sgid programs
> - several fd leaks due to missing close-on-exec flag
> - misspellings/typos in macro names in several headers
> - incorrect failure return value in inet_pton
> - various numeric ip address parsing and validation fixes
> - namespace conformance issues in several headers
> - minor header issues
> - zombie processes left by faccessat with AT_EACCESS
> - timezone file parser failing/crashing on 64-bit archs
> - timezone path search was only trying first path
> - incorrect handling of excessive-length TZ environment strings
> - timezone file loading was wrongly enforcing O_NOFOLLOW/rejecting symlinks
> - iswspace was wrongly returning true for the null character
> - various bugs in wordexp
> - putgrent could write corrupt lines after write failures
> - dn_expand misinterpreted in-packet offsets greater than 255
> - spurious strftime/wcsftime failure on len+1==bufsize case
> - incorrect underflow flag in fma corner cases
> - log*(0) wrongly returned +inf in downward-rounding mode
> 
> i386-specific bugs fixed:
> - failure of fesetround to set sse rounding mode
> - floating point limit constants interpreted wrong due to excess precision
> 
> powerpc-specific bugs fixed:
> - broken thread pointer access when compiled with clang
> 
> strict conformance issues:
> - NULL definition re-aligned with POSIX (requires (void *) cast)
> - alignment of math.h is* comparison functions with C11 annex F requirements