- IRIX: Get MAC address from ioctl(), not by spawning shell. (Thomas Habets)
- Compile fix for IRIX (added includes) (Thomas Habets)
- Don't explicitly check for UID 0, we may have capabilities even if not root. (Thomas Habets) (Sam Roberts)
- Cleaned up implementations of libnet_get_hwaddr(), some leaked memory, one returned a pointer to data on the stack, and the others return a pointer to static data. I'm settling on the non-reentrant static data form. (Sam Roberts)
- Visual C++ compiler(v9.0) uses bitfield type as a hint to pad the bitfield, so struct was too long. (Sam Roberts)
- Further simplify autogen.sh (Sam Roberts)
- Removed dependency on net/bpf.h, and on pcap.h. (Sam Roberts)
- LBL_ALIGN check is unused, removing. (Sam Roberts)
- Don't include pcap.h if we've already got net/bpf.h, pcap has it's own bpf. (Sam Roberts)
- Make libnet_get_hwaddr work in the last few releases of OpenBSD (stu@spacehopper.org) (Sam Roberts)
- Get DLT types from the source, <pcap.h>. (Sam Roberts)
- Use autoconf -ivf in autogen.sh (suggested by alon.barlev@gmail.com) (Sam Roberts)
- Don't distribute libnet.h, it is generated by configure (alon.barlev@gmail.com) (Sam Roberts)
- AIX build failures fixed, cause was inclusion of system headers libnet no longer uses (alon.barlev@gmail.com) (Sam Roberts)
- Add a link to the old docs. (Sam Roberts)
- Added links to github and sourceforge. (Sam Roberts)
- Replace C99/C++ comments with traditional C comments. (alon.barlev@gmail.com) (Sam Roberts)
- checksum would segfault if a IP checksum was requested for a non-ipv4 header (Sam Roberts)
- Closer backwards compat, assume its ipv4 if it's not ipv6. This seems totally wrong, but so it goes. (Sam Roberts)
- Try using the ip_len header field to guess the input buffer's size. (Sam Roberts)
- libnet_do_checksum(), despite being "internal", is used by external code. libnet needs to maintain backwards API compatibility, tcpsic from the isic package is an example of a binary failing when calling the new API with the old arguments. (Sam Roberts)
- merged icmpv6 patch in, but I believe either it or the sample/icmp6_unreach generates the cksum incorrectly (victor@inliniac.net) (Sam Roberts)
- Use libtool-1 or libtool-2 whatever available (alon.barlev@gmail.com) (Sam Roberts)
- Avoid looking at /dev and /usr/include when cross-compiling (alon.barlev@gmail.com) Alon: The following code support cross compiling: 1. You CANNOT check for /usr/include stuff as cross compiler is installed else-where. Autoconf know how to do this, use its header detection logic and ask the result. 2. Testing for /dev/ can be done only when not cross compiling... (Sam Roberts)
- Corrected target_os check, it was broke for linuxgnu, and m4 syntax was invalid (alon.barlev@gmail.com) Alon:The following change is needed in order to solve two issues: 1. linuxgnu and such target os. 2. You cannot set variable with space before '=' as it tries to execute the variable... 3. Print result of test in case of linux (Sam Roberts)
- --with-link-layer broken, was using wrong macro name, and didn't include all link types (Sam Roberts)
- ac_cv_c_bigendian is yes, not "big" (alon.barlev@gmail.com) (Sam Roberts)
- Remove dead code. (Sam Roberts)
- h_len is calculated for ip, udp, tcp, icmp, and igmp, so is allowed to be zero. (Sam Roberts)
- libnet_t's fd should be initialized to an invalid value, or libnet_destroy() will close stdin. (Sam Roberts)
- Alon's use of AC_CHECK_HEADERS fails to detect headers. Reverted part of 57acd56 (Sam Roberts)
- With link_none, the link apis were failing with not error message. (Sam Roberts)
- man doc makefile wasn't correctly referring to the srcdir (alon.barlev@gmail.com) (Sam Roberts)
- Avoid mallocing zero bytes, it perturbs electric fence. (Sam Roberts)
- autotools patches for cross compiling and seperate builddir (alon.barlev@gmail.com) (Sam Roberts)
- html doc makefile wasn't correctly referring to the srcdir (alon.barlev@gmail.com) (Sam Roberts)
- ip_offset is now calculated on the fly, and UDP and TCP no longer use h_len (Sam Roberts)
- IP offset calculation should allow nesting of IP protocols. (Sam Roberts)
- Remove gccisms in bitfield definitions. (Sam Roberts)
- injection type of LIBNET_NONE, for packet construction without injection (also, more const correctness) (Sam Roberts)
- Notes about checksumming. (Sam Roberts)
- libnet_build_tcp was not returning the ptag. (Sam Roberts)
- Packet boundaries are now passed to _do_checksum(), so it can validate its input. Hopefully, this will end the recurring segmentation faults due to buffer overruns. (Sam Roberts)
- TCP building is triggering memory overwrites; closer examination shows the link list manipulation to be wrong, and the checksumming approach to be incapable of working. I reworked code to simplify and clarify how it works currently, in preparation to fixing it. (Sam Roberts)
- Updated comments and notes. (Sam Roberts)
- Null the pointer in the about-to-be-freed structure, not the one on the stack. (Sam Roberts)
- Added missing pblock types, and made strings consistent with definitions. (Sam Roberts)
- Change version policy, we will be 1.1.4 until 1.1.5 is released. (Sam Roberts)
- Bring CHANGELOG up to date with today, and script used to generated it. (Sam Roberts)
- Begin implementation and tests for repairing pblocks after an update. (Sam Roberts)
- libnet_pblock_insert_before() didn't remove ptag2 from old location (Sam Roberts)
- Clarifications in document comments. (Sam Roberts)
- Don't doxygen process internal header libnet-headers.h (Sam Roberts)
- Declared many constant arguments as const, const-correct code spews warnings when built against libnet. (Sam Roberts)
- Include pcap DLT_ types from correct header, was using an internal one before. (Sam Roberts)
- Declared many constant arguments as const, const-correct code spews warnings when built against libnet. (Sam Roberts)
- Note about build_data, which doesn't update ip_offset, among other problems. (Sam Roberts)
- libnet_clear_packet() wasn't clearing all packet context. (Sam Roberts)
- Why don't TCP and UDP use the DATA pblock type? (Sam Roberts)
- whitespace cleanup (Sam Roberts)
- Add libnet_dll.c as extra, so its there for win32, and build libnet_link.c (Sam Roberts)
- This file wasn't being built, and needed to include bpf to build. (Sam Roberts)
- Forgot to make device a const string here, too. (Sam Roberts)
- make string argument constant (Sam Roberts)
- only ignore Makefile in libnet/ (Sam Roberts)
- Replace u_intX_t with C99 uintX_t. (Thomas Habets)
- pclose() following popen(), not fclose() (Thomas Habets)
- snoof & dlpi: don't free on libnet_link_close() (Thomas Habets)
- Summarize changes for log. (Sam Roberts)
- Reindented, removing hard tabs, and using consistent brace positioning. (Sam Roberts)
- The non-standard types are no longer used. (Sam Roberts)
- /sw/.. path doesn't always exist (Sam Roberts)
- src/libnet_link_snoop.c: Only fclose if f!=NULL (Thomas Habets)
- src/libnet_link_snoop.c: fixed snoop-based backend. Works on IRIX. (Thomas Habets)
- Use uint64_t, not u_int64_t (Thomas Habets)
- define a lying gethostbyname2() if it's not defined (Thomas Habets)
- define STDOUT_FILENO if it's not defined (Thomas Habets)
- configure.in: check for gethostbyname2 (Thomas Habets)
- configure.in: Check for uint{16,32,64}_t (Thomas Habets)
- Configure switch to install samples (Sam Roberts)
- Attempt at applying a patch to get installable samples, which doesn't work. (Sam Roberts)
- Pointers not cleared after free could lead to double deallocation. (Sam Roberts)
- Convert CRLF to LF. (Sam Roberts)
- Auto* changes to work on OS X from git checkout. (Sam Roberts)
- Update autobuild endianness and unaligned checks. (Mike Frysinger) (Sam Roberts)
- Add srcdir to include path. (Sam Roberts)
- Adjust srcdir and builddir so libnet can build out-of-tree (Robin Getz/Mike Frysinger) (Sam Roberts)
- Beginning 1.1.5 development. (Sam Roberts)