The beta4 of WinPcap 3.1 is available from today in the download section of
the WinPcap site.
WinPcap 3.1 beta4 is a minor update, that fixes several bugs and leaks
reported by our users after the release of beta2 and beta3.
Changelog
=========
- wpcap.dll has been updated to libpcap 0.8.3 from http://www.tcpdump.org.
- Added a note in the documentation that states that the kernel dump feature
is disabled due to incompatibilities with the new kernel buffer.
- Minor fixes to the documentation
- Removed some useless files.
- Bug fixing:
+ Fixed a bug related to COM initialization in WanPacket, by which
WanAdapters were not working correctly if the calling thread was using
COM with a different threading model.
+ Fixed a problem in AddAdapterIPH(), by which no adapter was actually
added with this function because of a UNICODE/ASCII mismatch. Basically,
AddAdapterIPH received an ASCII adapter name, and tried to open it with
PacketOpenAdapterNPF, which accepts UNICODE strings, only.
+ Fixed a bug in the remote capture code due to concurrency issues when
spawning a new thread
+ Fixed a problem related to the generation of grammar files with flex
in the CygWin makefile.
+ Fixed a couple of memory leaks in PacketGetAdapterNames().
PacketGetAdapterNames() seems to be still leaky, but the source of the
leak seems to be a leaky API in the Microsoft IpHelperAPI, at least on
WinXP SP1.
+ Added some code that frees the global list of adapters when packet.dll
is unloaded (i.e. when DllMain is called with DLL_PROCESS_DETACH)
+ Fixed a bug that caused the adapters not to be listed on terminal
services. The bug was caused by the lack of the "\\global" prefix in
front of the adapter names.
+ Fixed a bug related to adapter opening in the pcap_filter example. Fixed
the usage string that was wrong.
+ Fixed a bug in the JIT code of the driver that could potentially cause a
BSOD if two threads try to set a filter (that will be jitted) at the
same time.
+ Fixed a bug by which the driver fails to return any packet with a read
after an IOCTL_SETBUFFER has changed the buffer size. The bug is due to
some missing counter resets.
+ Fixed some debugging messages in the NT driver that were not macroed
with IF_LOUD
=========