0.31:
special case buffer_get_token with token length 1 through memccpy (almost 4x speedup)
test for pread and use it instead of lseek+read in cdb and io_mmapwritefile
add mmap_readat (like mmap_read but open by fd, not by name)
add fmt_iso8601 and scan_iso8601 (time_t -> "2014-05-27T19:22:16Z")
add fmt_netstring and scan_netstring (http://cr.yp.to/proto/netstrings.txt)
add (fmt|scan)_varint and (fmt|scan)_pb_* (binary marshaling ala protocol buffers)
0.30:
add compiletimeassert.h
add haveuint128.h
byte.h now defines UNALIGNED_ACCESS_OK on x86
add scan_ulongn, scan_8longn, scan_xlongn, scan_longn, scan_netstring
change buffer to have a destructor function pointer
SECURITY: fmt_strn would write one byte too many (returned right length though, so usually not a problem as that byte would be overwritten with \0 by the caller anyway)
fmt_pad and fmt_fill fail more gracefully when srclen > maxlen
You can now say $ make WERROR=-Werror (compiling t.c may fail but that's strictly speaking not part of the library)
scan_html now decodes HTML entities based on the actual w3c list (from entities.json, say gmake update to fetch the current version)
added fmt_escapechar* to fmt.h (implement various escaping mechanisms also found in textcode but for a single char not a whole string, and they always escape, not just when they think it's needed)
scan_ushort was supposed to abort early and return 5 when attempting to parse "65536", because the result does not fit. It did not. Now it does.
scan_*long, scan_*int, scan_*short now properly abort if the number would not fit
SECURITY: check for integer overflow in stralloc_ready
switch io_fds from array to newly implemented (hopefully thread-safe) iarray
switch epoll from level triggering to edge triggering
introduce io_eagain_read and io_eagain_write (discontinue using io_eagain plz)
fix buffer_get
add fmt_html_tagarg, fmt_xml, scan_html_tagarg
add socket_fastopen, socket_fastopen_connect4, socket_fastopen_connect6 and socket_quickack
on Linux, in iob_send, if MSG_MORE is defined, and the request type
lends itself to it, used sendto/sendmsg with MSG_MORE instead of
TCP_CORK (saves two syscalls)
add io_fd_flags so the caller can tell io_fd whether the socket is blocking
(saves one fcntl syscall)
more constness for stralloc and buffer
mmap_read/mmap_shared on zero length files no longer fail but return a
zero length buffer
if SOCK_NONBLOCK is defined, use it instead of socket+fcntl
... but if errno==EINVAL still fall back to socket+fcntl (Robert Henney)
SECURITY: fix botched integer overflow handling logic in stralloc_ready (Giorgio)
add critbit
0.29:
save 8 bytes in taia.h for 64-bit systems
add buffer_tosa (buffer writing to auto-growing stralloc)
add iarray
support SCTP in addition to TCP
fix io_receivefd so the incoming buffer is only 1 byte; io_passfd
sends only one byte, so we might receive (and discard) other data if
we try to read more (Stefan Bühler, from the lightttpd project)
add io_fd_canwrite (like io_fd but assume the fd is writable)
save a few syscalls here and there
add awesome hack in isset.h (comex, Linus Torvalds)
add byte_equal_notimingattack
Issues found with existing distfiles:
distfiles/eclipse-sourceBuild-srcIncluded-3.0.1.zip
distfiles/fortran-utils-1.1.tar.gz
distfiles/ivykis-0.39.tar.gz
distfiles/enum-1.11.tar.gz
distfiles/pvs-3.2-libraries.tgz
distfiles/pvs-3.2-linux.tgz
distfiles/pvs-3.2-solaris.tgz
distfiles/pvs-3.2-system.tgz
No changes made to these distinfo files.
Otherwise, existing SHA1 digests verified and found to be the same on
the machine holding the existing distfiles (morden). All existing
SHA1 digests retained for now as an audit trail.
This changes the buildlink3.mk files to use an include guard for the
recursive include. The use of BUILDLINK_DEPTH, BUILDLINK_DEPENDS,
BUILDLINK_PACKAGES and BUILDLINK_ORDER is handled by a single new
variable BUILDLINK_TREE. Each buildlink3.mk file adds a pair of
enter/exit marker, which can be used to reconstruct the tree and
to determine first level includes. Avoiding := for large variables
(BUILDLINK_ORDER) speeds up parse time as += has linear complexity.
The include guard reduces system time by avoiding reading files over and
over again. For complex packages this reduces both %user and %sys time to
half of the former time.
