Changes based on ndis-1.2.6 (author uncertain, Ken Mays?)
Note that this is geared heavily toward Broadcom WiFi support (64-bit NDIS5
tested with certain releases of binary drivers)

================ ndis.1.3.0, Jun 17, 2013
- A few cosmetic fixes

================ ndis.1.3.0.rc3, Apr 27, 2013 (Jim Klimov)
- Makefile enhancements (automate build, install and test, parametrize things)
- Makefile.com enhancements (reference gcc444, set NDIS=bcmndis)
- Some source code fixes to reduce build warnings (i.e. variable types,
  include files)
- Added __amd64__ paddings into structure field alignments (ntoskrnl_var.h)
  and changed layout_check() to be a more verbose series of ASSERT_EQ's
- Added a contrib/wifi - an init-script that can configure and plumb the
  wireless interface (or unconfigure/unplumb/modunload) as both help during
  tests and as a precaution against a driver that would kernel-panic quickly
  (includes a block-file to not load a failed driver upon next boot; possibly
  you should install the drivers into /usr/kernel/{misc,drv}/amd64/ then, so
  that they are not automatically loaded by the OS itself)
- Moved an empty but required restore() function from C to ASM so that it is
  not optimized out (JPA's idea, good hit)

================ ndis.1.3.0.rc2, Apr 23, 2013 (Jean-Pierre Andre)
- Implemented a mutex in KeAcquire/KeReleaseSpinLockFromDpcLevel
- Added Windows struct layout checks
- Cleaned some confusion between ndis_softc and ieee80211com
- Referenced the implemented getprop/setprop instead of the stubs
- Enabled partial Brussels (WPA) support
- Initialized the ieee80211 library
- Fixed the trace display of IoGetDeviceObjectPointer()
- Added stubs for functions required by newer drivers
- Implemented ExFreePoolWithTag()
- Made a real return from IoAllocateIrp()
- Improved the traces for delayed replies in getprop/setprop
- Protected against device_object poorly initialized
- Removed obsolete functions
- Added missing local prototypes
- Avoided ndis_handle to mean handle to ndis_miniport_block
- Avoided ndis_handle to mean handle to ndis_packet
- Reworked the simulation of IRQ levels
- Used space for alignment instead of superimposing fields
- Fixed delays for waiting for an event
- Put lengths to return for getinfo/setinfo in permanent storage
- Serialized the requests for getinfo/setinfo
- Set the queues before the workers start
- Implemented a specific thread for clock ticks
- Made to wait until next clock tick when driver is busy
- Returned a proper error when an unknown stat is requested
- Removed obsolete logging

================ ndis.1.3.beta6, Mar 26, 2013 (Jean-Pierre Andre)
- Fixed references to included files
- Avoided a duplicated external __progname
- Redefined strcasestr to be Posix compliant
- Changed the way the ndis_regvals table is generated
- Fixed unmatching types
- Fixed the definition of type quehead
- Defined wifi_bcastaddr external
- Avoiding alignment requirements in cm_partial_resource_desc and io_stack_location
- Logged the location of driver for easier debugging
- Removed calls to printf from the driver
- Avoided double fclose() in ndiscvt
- Removed unused variables
- Fixed wrong signed vs unsigned types
- Avoided exporting data
- Fixed arguments relocations in x86_64_call5()
- Included borrowed code for mi_strlen and mi_mpprintf_putc
- Used the same trace code for subr_pe in user and kernel mode
- Shared more wrapping code for easier debugging
- Avoided panicking when disabling the driver which failed to load
- Relocated arguments to functions called by ntoskrnl_run_dpc()
- Inserted a normal return into main() of ndiscvt
- Made arguments in Windows mode for ndis_timercall()
- Used macroes to enable linking to the driver at build time
- Made the Windows driver linkable at build time
- Reserved at least 32 bytes in the stack for calling Windows-type functions
- Cosmetics
- Configured a void private property list
- Protected against sending single packet not being defined
- Avoided inserting noise into the linked driver
- Added a check on the Windows driver size
- Made sure the driver is correctly registered
- Secured a check on the driver registration
