A new test that triggers the bug (with ARPSourceMACField) has been added.

Fix IPv6 support on Windows

  - LLTDAttribute() produces a Raw instance (or crashes, when
    conf.debug_dissector is set).

  - str(LLTDAttribute(type=18)) crashes.

Thanks @gpotter2

Pypy won't run crypto tests on Travis CI because the cryptography
module cannot be installed with the Pypy version used.

Change the cryptographic library used in the following layers:
 * IPSec
 * TLS
 * Dot11

Notes:
 * Fix the WEP key in regression.uts: Should be a 40 bits key (5 ASCII)
   and 24 bits nonce for 64 bits WEP. The key was 8 ASCII => 88 bits
   in total
 * AES-GCM and AES-CMAC algorithms are now native in cryptography: add
   tests in ipsec.uts
 * cryptography only supports Python 2.6+
 * Fix the docstring for the pad method

Removed:
 * MD2 and MD4 hash algorithms for RSA in tls/crypto/pkcs1.py
 * AES-XCBC-96 not supported by cryptography.

Signed-off-by: Thomas Faivre <thomas.faivre@6wind.com>

One can now use tcpdump or tshark to dissect packets from
Scapy. This can help, for example, to check Scapy's dissectors

This fixes at least an issue fragmentation (for which a new test
has been added)

* Add conf.checkIPinIP

This setting, when set to False, will exclude IP layers which encapsulate
another IP layer from the checks to tell if a packet is an answer to
another one.

Fixes #383

* conf.checkIPinIP affects both IP() and IPv6()

Setting conf.checkIPinIP will handle IP()/IP(), IP()/IPv6(),
IPv6()/IP(), IPv6()/IPv6().

Thanks @guedou!

* Clean-up IP.hashret and .answers code

* Add tests for .hashret() and .answers()

* Support (BPF) filter in sniff() with offline parameter set

Fixes #393
Also, fixes #355

* Add tests for wrpcap() and sniff(offline=)

as suggested by Guillaume.

Also, cleanup regression.uts since it was a pain to find a place
to add those tests.

* Fix PATH for tcpdump with non-root user

* Do not run tcpdump tests when tcpdump is not available

* Appveyor tests: install WinDump.exe

Thanks @gpotter2

- original patch from Yitai Schwartz, https://bitbucket.org/secdev/scapy/pull-requests/62/

The "NextProtocol" field was added as an unconditional field which made the
VXLAN packet header consume an additional 8-bts when used in a Legacy VXLAN
deployments (flags=0x08) or in GPID deployments (flags=0x88).  The resulting
regression caused incorrect parsing of subsequent headers.  For example, if an
Ethernet header followed the VXLAN header then it would not be interpreted as
an Ethernet header.  The actually interpretation would depend on what value was
stored in the Ether.dst field.

This change updates the field layout specification such that the NextProtocol
field is conditional on flags=0x04, and shortens the reserved field by 8-bits
if that condition is true.

fixes bug #281

Signed-off-by: Allain Legacy <allain.legacy@windriver.com>

Testing:
Updated dhcp6 regression tests

Since part of the previously reserved Bytes are now used to annouce
the next header, this also updates the vxlan unit tests.

Original commit by: Christophe Fontaine <christophe.fontaine@qosmos.com>

Signed-off-by: Gabriel Ganne <gabriel.ganne@qosmos.com>