ReleaseNotes

      - examples/udp:  Port numbers need to be configurable to work with
        6LoWPAN.  Need to bind the client socket to a port number.  This was
        not required before so is a apparently a change in the UDP packet
        dispatch logic.
      - examples/udp:  Enable testing with the broadcast address.
      - examples/nettest:  Support target boards on both ends of the test.
        Separate out network initialization so that it may be used by both a
        target server and a target client.  Fix client/server naming
        confusion; Add command line option to select the server address on the
        target.  Add support for both enpoints on target boards vs. one on a
        target and one on the host PC.
      - examples/nettest:  The send buffer size is now a configuration option.
      - examples/nettest:  Loopback option should be available in Kconfig for
        PktRadio.
      - examples/keypadtest: REMOVED and warehoused in the Obsoleted
        repository.  This was just a bad clone of apps/examples/hidkbd for a
        keypad driver that was removed years ago.  It also uses illegal
        function calls into the OS.  So it has no purpse:  It is redundant, it
        uses illegal interfaces, and is a test for non-existent code.
      - examples/ipforward: Add an IP forwarding example using only TUN
        devices.  Test extended to test forwarding of ICMPv6 multicast messages.
      - Implement powermonitor example for ltc4151 current and voltage
        monitor.  From Giorgio Groß.
      - examples/pf_ieee802154:  Add for testing PF_IEEE802154 sockets.  Add
        PANID to command line options; Cannot bind to address zero... There is
        no counterpart to INADDR_ANY for these radios (not now at least).
      - examples/dac: add DAC example.  From Juha Niskanen.

    * Network Utilies: apps/netutils:

      - netutils/telnetc:  This a port of libtelnet to NuttX.
      - netutils/telnetd and users of telnetd:  Add support for IPv6.
      - netutils/thttpd:  Remove the netstats demo.  This depends on an
        illegal function call and cannot be supported.  That example could be
        replaced with logic that uses the procfs network entries as was done
        for NSH which had the same issue.  But I am too lazy to implement that.
      - netutils/netlib:  Add support for PktRadio IOCTL commands.
      - netutils/netlib:  Add a helper to decode short addresses.

    * Wireless Utilies: apps/wireless:

      - wireless/ieee802154/i8sak:  Rewrote i8sak to be test CLI for IEEE
        802.15.4 MAC layer.  From Anthony Merlino.
      - wireless/ieee802154/i8sak:  Updates to i8sak for
        association/beacon-enabled changes to IEEE 802.15.4.  From Anthony
        Merlino.
      - wireless/ieee802154/i8sak:  Add option to make it easy to send large
        frame for testing purposes.  From Anthony Merlino.
      - wireless/ieee802154/i8sak:  Add scan command.  From Anthony Merlino.
      - wireless/ieee802154/i8sak:  Add coordinfo command allowing you to
        poll various attributes related to the coordinator.  From Anthony
        Merlino.
      - wireless/ieee802154/i8sak:  Add a option to retry on failures to the
        assoc command (usually meaning tht the coordinator is not yet up).
      - iwireless/ieee802154/8sak:  Add reset command to reset the MAC layer.
        Adds option to assoc command -t to specify how long to wait for a
        response from the Coordinator.  From Anthony Merlino.
      - ieee802154/i8sak:  Adds command to trigger regdump of radio.  From
        Anthony Merlino.

    * System Unitilities (apps/system)

      - apps/system/composite: Update to apps/system/composite assocated with
        big changes to the composite device logic.  From Frank Benkert.
      - apps/system/composite:  Remove CDC/ACM and MSC configuration logic.
        This belongs in the OS composite initialization.  Add and argument so
        that you can select the USB composite configuration to be attached.
        Restore USB tracing; remove unused field in a structure.  Remove
        configuration settings that are no longer used.  Fix configuration
        selecting.. was setting the port number, not the configure ID.  Also
        add a configuration option to select the default configuration.
      - apps/system/composite: Add a configuration option to the boardctl()
        calls to support multiple composite device configurations dynamically.
      - apps/system/composite:  Remove references to USBMSC.  There still
        dependencies on CDC/ACM in the serial USB trace output.
      - apps/system/telnet:  Add Telnet Chat deamon and and client from
        libtelent.

    * Platform-Specific Support (apps/platform)

Bugfixes.  Only the most critical bugfixes are listed here (see the
ChangeLog for the complete list of bugfixes and for additional, more
detailed bugfix information):

    * Core OS:

      - Fix ELF loader up_checkarch on ARM arch.  From Cristian Condurache.
      - pthread_mutex_unlock():  Missing check for the case where
        pthread_mutex_lock() is called when the mutex is not locked.  In that
        case, it would increment the underlying semaphore above 1.  This is
        the fix for a problem noted by initialkjc@yahoo.com.
      - sig_timedwait():  Pending signal structure used after it has been
        releasd.  From anonymous Bitbucket Issue 59.
      - mm_mallinfo:  do heap end debug assert check with heap semaphore held.
        From Jussi Kivilinna.
      - sched/:  Fix return value in sched_setaffinity().  From Masayuki
        Ishikawa.
      - Work Queues:  work_queue() must cancel existing work prior to queuing
        new work, otherwise the work queue list structure can become
        corrupted. Problem noted by Pascal Speck.
      - waitpid():  Corrects two problems when CONFIG_SCHED_HAVE_PARENT +
        CONFIG_SCHED_CHILD_STATUS are enabled:  (1) Was erroring out if the
        waited for task had already exited, and (2) was not freeing resources
        when a wait was completed.  From Boris Astardzhiev.

    * File System/Block and MTD Drivers:

      - VFS poll():  Fix timeout calculation.  From Jim Paris.
      - VFS poll():  Fix poll for regular files and block devices.  Open Group
        documentation says that poll (and select) support regular files and
        that 'Regular files shall always poll TRUE for reading and writing'.
        From Jussi Kivilinna.
      - VFS epoll():  Fix epoll_wait function.  From Simon Piriou.
      - Smart FS:  Fix wrong freeing of device structure and use-after-free
        issues on error paths.  From Jussi Kivilinna.
      - MTD FLASH drivers:  The byte write method of several drivers had a
        cloned error:  It was not locking the bus while performing byte write
        operations.
      - MTD M2PX: If we READ while a write/erase is pending, the command is
        ignored and the write/erase continues. If we dont catch this situation
        we will return garbage to the user because the flash will not execute
        the command.  So READ MUST wait for write completion, and before that,
        the bus must be locked since it's a precondition to calling
        waitwritecomplete().  From Sebastien Lorquet.
      - MTD FLASH drivers: Clone Sebastien Lorquet's m25px change to at25,
        is25xp, ramtron, and sst25xx.
      - MTD W25:  Add missing locking and fix SPI_SELECT usage for
        w25_unprotect.  From Jussi Kivilinna.
      - MTD W25:  Wait for BUSY flag to clear in w25_readid and
        w25_unprotect.  W25Q128 datasheet says that all instructions expect
        'Read Status Register' and 'Erase/Program Suspend' are ignored when
        BUSY flag in status register is '1'. Therefore wait for busy flag to
        clear in w25_readid() and w25_unprotect().  From Jussi Kivilinna.
      - Automounter: FS_AUTOMOUNTER should depend on SCHED_LPWORK.  From
        Nickolay Semyonov.

    * Networking/Network Drivers:

      - TCP Listen: Throw error when error happens in the tcp_listen
        function.  From Roland Takacs.
      - Nework Device Management:  Do not search net device when all-zeros
        address is used.  From Roland Takacs.
      - Network Device Management:  Fix a error in netdev_register(); it was
        not handling device names properly when TUN is the only network device.
      - Network Device Management:  Fix netdev_dev_lladdrsize().  In some
        configurations, it could return the wrong size for the address of a
        packet radio.
      - Network Device Management:  Fix typo for 802.11 devices in
        netdev_register().  Was being masked before because depended on
        CONFIG_NET_MULTLINK.
      - TUN Driver:  Use critical section instead of semaphore in
        tun_ifdown().  From Masayuki Ishikawa.
      - TUN Driver:  Do all polling on worker thread.  Otherwise, the stack
        gets very deep.
      - TUN Driver:  Currently cannot support TAP mode unless Ethenet is
        enabled.
      - netdb:  Fixed buffer size used for sending DNS queries should depend
        on the configured DNS name size.  From Ritjaina.
      - UDP networking:  The TTL (time to live) was not being set in the IPv4
        or IPv6 header unless the UDP socket was bound.
      - UDP Networking:  Fix a copy-paste error that could effect networking
        when both IPv4 and IPv6 are enabled.
      - UDP Networking:  When upd_input() cannot process a packet, it returns
        ERROR so that network drivers may try calling ipv4_input() later.  In
        this case, it must also set d_len to zero.  Otherwise, all network
        drivers will assume tht there is also an outgoing packet.  This
        results in a gratuitous ARP.
      - TCP Networking:  Correct some issues that prevent TCP from working
        correctly when both IPv4 and IPv6 are enabled.
      - TCP Networking:  TCP disconnection callbacks are not retained in a
        list.  This will support mutiple callbacks per lower-level TCP
        connection structure.  That is necessary for the cae where a socket is
        dup'ed and shares the same lower-level connection structure.  NOTE:
        There still needs to be a call to tcp_start_monitor() when the socket
        is dup'ed.
      - TCP Networking:  Start the network monitor for a socket when a TCP
        socket is dup'ed.
      - TCP Networking:  If one of the dup'ed socket's is closed, then network
        monitor resources associated with that one socket must be recovered.
        Also, in the event that socket is being used on one thread, but then
        closed on another, any threads waiting for events from the socket
        should be informed of the closure.  That latter requirement is not
        implemented because current data structures do not support it.
      - TCP Networking:   Fix a race condition.  The accept() operation is
        performed with the network locked.  However, the network is unlocked
        BEFORE the connected state is set.  Therefore, a context switch may
        occur and the socket may no longer be connected when it is marked so.
        Noted by Pascal Speck.
      - Network routing tables:  Fix a compilation error when IPv6 and
        routing are enabled.
      - Network procfs:  Fix some spacing when both IPv6 and IPv4 are enabled.
      - Network Local Sockets:  Fix accept for local stream sockets.  From
        Jussi Kivilinna.
      - Network Local Sockets:  Fix server lc_waitsem overflow.  From Jussi
        Kivilinna.
      - IPv6 Networking:  Remove comparisons to the address with all ones
        set.  IPv6 does not support broadcast addresses and certainly not in
        that form.  Replace with multicast addresses beginning with 0xff02.
      - ICMPv6 Networking:  Fix a compilation issue with
        CONFIG_NET_ICMPv6_AUTOCONF=y.  From  Masayuki Ishikawa.
      - Networking:  Fix net_lock returning ERROR when instead of real error
        code on failure.  From Jussi Kivilinna.
      - IGMPv2 Networking:  Remove special support for interrupt level
        processing (there is none) and fix some timer cancellation logic.  In
        many files, correct comments.  There is no interrupt level processing
        in the networking layer.

    * Wireless/Wireless Drivers:

      - IEEE 802.15.4 MAC:  Need counting protection on the logic that
        releases the notification resources.  Otherwise, notification handlers
        may be operating with a stale pointer.
      - IEEE 802.15.4 MAC: Improves internal timer logic to handle work
        serially.  Before, the MAC timer used a watchdog to schedule work with
        the high priority worker queue.  However, since everything in the MAC
        is supposed to be serialized through the use of the high priority work
        queue, but the timer uses a watchdog, there are some unintended
        consequences. To simplify, we now use the delayed work feature of the
        work queue.  From Anthony Merlino.
      - IEEE 802.15.4 Network Driver: Update RX statistics in network driver.
      - IEEE 802.15.4 MRF24J40 Driver:  Minor timing fix. Matches recommended
        value in datasheet.  Splits up driver into multiple files to make it
        easier to navigate.  Fixes issue with non-beacon enabled mode.  From
        Anthony Merlino.
      - IEEE 802.15.4 MRF24J40 Driver:  Fix a bug causing radio to cease
        transmitting.  From Anthony Merlino.
      - IEEE 802.15.4 MRF24J40 Driver:  Fixes issues with sleeping for beacon
        enabled networking.  From Anthony Merlino.
      - EEE 802.15.4 MRF24J40 Driver:  Fix math error for calculating sleep
        count values.  From Anthony Merlino.
      - IEEE 802.15.4: Fix timing issues for beacon enabled coordinators and
        endpoints. From Anthony Merlino.
      - iIEEE 802.15.4: Fix issue with association on beacon-enabled
        networking.  From Anthony Merlino.
      - 6LoWPAN:  Correct a few addressing issues.  Also reserve two bytes at
        the end of the frame for the FCS.
      - 6LoWPAN:  Fixes needed when extended addressing is enabled; broke short
        addressing.  Loopback driver needs to initialize the MAC meta data;
        Address decompression logic must have the MAC address to handle the
        most common compression cases.  Fix a src/dest address confusion and
        other addressing problems.
      - 6LoWPAN:  Add missing IPv6 address creation to HC1 decode logic. Fix a
        typo in an index that prevented use with HC1 and extended addresses.
      - 6LoWPAN:  TCP logic was not obeying MTU packet size limitations.
      - 6LoWPAN:  Major re-architecting of TCP logic to properly handle TCP
        stuff like ACKs and TDP windowing which were not properly covered in
        the initial design.
      - 6LoWPAN:  HC06, copy TCP header as though it were data.  TCP packet
        reassembly now seems to work OK.
      - 6LoWPAN:  Fix duplicate and bad memcpy in loopback driver.
      - 6LoWPAN:  Fix a misconception about HC06 16-bit IPv6 address
        compression.
      - 6LoWPAN:  TCP send logic was returning a failure in one case when, in
        fact, the send was successful.

    * Common Drivers:

      - USB MSC:  Use struct instead of pointer to the struct as sizeof
        argument in memset in usbmsc.c.  Otherwise it leads to error: argument
        to sizeof in memset call is the same pointer type struct usbmsc_lun_s
        * as the destination.  From Oleg Evseev.
      - USB MSC:  Add missing logic to define endpoints.  The composite
        changes broke the the non-composite, USB MSC only case because it omitted the critical setup when USB MSC was not part of the composite.
      - USB CDC/ACM.  Fix several known problems resulting from merge of USB
        composite device.  That merge now breaks some of the non-composite USB
        devices.
      - Button Driver:  Interrupts weren't enabled since nothing updates them
        after btn_poll() marks the file descriptor structure as being polling.
        From Jan Pobrislo.
      - ADC Driver:  Fix some data alignment issues in the ADC driver.
      - I2C Drivers:  Handle I2C_TRANSFER return value consistently.  Some I2C
        peripherals transfers return zero on success, others number of
        completed transfers.  Make drivers robust against this.  From Juha
        Niskanen.
      - COMP Driver:  Fix compilation errors when poll disabled.  From Mateusz
        Szafoni.

    * Simulation:

      - Simulation:  Fix mkdir issue in GNU target.  From Simon Piriou.
      - Simulation:  x86 stack needs to be aligned to 16-byte boundaries.
      - Simulation:  Fix building 32-bit simulation on 32-bit X86.  From Jussi
        Kivilinna.

    * ARM:

      - ARM: The older ARM7 and ARM9 configurations were determining CFLAGS
        based on the GCC version 4.x.x or not.  That needx to be extended for
        5.x.x and 6.x.x which also behave like 4.x.x.

    * ARMv7-M:

      - Fixed ARMv7-M Toolchain setting.  Cortex-M4 only have Single
        Precision FPU.  From Hidetaka Takano.
      - ARMv7-M syscall logic:  Clear bit 0 in PC settings.  Bit 0 is the
        thumb mode indication and should not be set in the PC.  This extra bit
        has not caused problems in the past, but seeing it set in the PC is
        unnerving.

    * Expressif ESP32:

      - Fix ESP32 gpio enable reg and default UART pin. Modify default UART
        pin for ESP-WROOM-32. Fix gpio enable reg.  From Sungki Kim.

    * Microchip/Atmel SAMD21 Drivers:

      - SAMD21:  Fix some SPI-related issues.  From Matt Thompson.
      - SAMD21 SPI: I was having issues with the bus freezing up .. slaves
        holding SDL low.. so I rewrote a good portion of the interrupt logic
        based on the application notes from Atmel.  One major improvement is
        using the RXNACK flag in the STATUS register, which indicates that no
        device responded to an address packet. Assuming that the chip will
        always give an interrupt status, I believe it's possible to eliminate
        the timer as well.  From Matt Thompson.
      - SAMD/L21:  Need to preserve errno value across syslog() call.
      - SAMD21:  Changes needed to get USB working.  From Matt Thompson.

    * Microchip/Atmel SAMv7 Drivers:

      - SAMv7 TWIHS:  TWIHS driver add reference counting. From David Sidrane.
      - SAMv7 CAN:  We discovered a problem with the samv7 mcan driver which
        results, under some circumstances, in a very high CPU load.  The
        problem occurs, and is easily reproducible, if the device is connected
        to a CAN network with a wrongly configured CAN speed (baud rate). In
        our tests we set the CAN speed of the device to 1000000 and the speed
        of the other CAN nodes to 500000. The device is restarted and sends a
        CANopen "bootup message" to the CAN network. This results in huge
        amount of errors messages on the CAN bus, probably because of the CAN
        feature for acknowledging error messages. The error messages can’t be
        read by the device because of the misconfigured CAN speed, instead the
        CAN chip reports lots of errors, which are reported to the application
        which uses the CAN driver (CONFIG_CAN_ERRORS is enabled).  The CAN
        errors are reported from the CAN chip via interrupts and thus the
        interrupt load is very high in this scenario. To fix the problem the
        driver now disables each RX error interrupt after it is occurred.  The
        RX error interrupts are turned back on if at least one CAN message is
        received successfully.  From Stefan Kolb.
      - SAMv7 CAN:  I discovered while working on the SAMV7 mcan driver that
        the implementation of the CAN error handling is suboptimal. In the
        current implementation the many errors are implemented as pending
        errors.  But those errors are not pending, the errors occurred and are
        gone directly afterwards. This commit changes the described behavior
        and simplifies the handling of CAN errors.  From Stefan Kolb.

    * Microchip/Atmel SAMv7 Boards:

      - SAMv71-XULT:  Fix MRF24J40 interrupt GPIO number.

    * NXP/Freescale Kinetis:

      - Kinetis MPU:  Disable MPU when not in protected mode.  The hardware
        reset state of the the MPU precludes any bus masters other then DMA
        access to memory. Unfortunately USB and SDHC have there own DMA and
        will not have access to memory in the default reset state.  This change
        disabled the MPU if present on system startup.  From David Sidrane.
      - Kinetis MPU:  Fixed warning for kinetis_mpudisable. Missing header
        file added.  From David Sidrane.
      - Kinetis SIM:  Ensure isolation of clock dividers for 0 value case.
        This fixes a bug were a SoC does not have a clockdivN register and
        passes a 0 for the init value.  This prevents overflow of the 0
        decremented to -1 (0xffffffff) spilling over to other clockdivN
        fields.  From David Sidrane.

    * NXP/Freescale i.MX6 Drivers:

      - i.MX6: Fix a wrong parameter passed when calling irq_attach() in
        imx_serial.c.  From Masayuki Ishikawa.
      - Based on Masayuki's change, I review all serial driver vector
        attachment.  I Found one additional error and updated all relevant
        drivers to current interrupt parameter passing.

    * STMicro STM32:

      - STM32 F410 Kconfig:  Fix peripherals available on the STM32 F410.
        This also adds a select for STM32_HAVE_DAC1 present on this STM32
        flavor.  From Gwenhael Goavec-Merou.
      - STM32 L4 DMA: Correct USART3_RX bad channel definition.  From Juha
        Niskanen.
      - STM32 L4 PWR:  Correct PWR_SR2 REGLPS and REGLPF bits, add port I
        registers.  Also remove duplicate section from Kconfig.  From Juha
        Niskanen.
      - STM32 F7: do not enable read-modify-write on DTCM.  "AN 4667 - STM32F7
        Series system architecture and performance" recommends to disable
        read-modify-write on DTCM: "If the DTCM-RAM is used as data location
        and the variables used are byte or/and halfword types, since there is
        no ECC management in this RAM on the STM32F7 Series, it is recommended
        to disable the read-modify-write of the DTCM-RAM in the DTCM interface
        (inthe DTCMCR register) to increase the performance." From Jussi
        Kivilinna.

    * STMicro STM32 Drivers:

      - STM32 TIM3/4 are always 16-bits; never 32-bits.  Noted by Eetu
        Nevalainen..
      - STM32 ADC:  Invalidate dma buffer before use.  Missing invalidation
        caused old samples being fetched from cache.  From Jussi Kivilinna.
      - STM32, STM32L4, STM32F7 ADC:  Fix channel 18 sample time.  From Juha
        Niskanen.
      - STM32 DAC:  Fix some configuration logic. When STM32_NDAC is greather
        than 1, then second channel is always DAC1OUT2.  From Mateusz Szafoni.
      - STM32 DAC:  Fix compilation when DMA disabled for channel.  From
        Mateusz Szafoni.
      - STM32 F0:  Fix some funny shifts in DAC header files.  From Juha
        Niskanen.
      - STM32 F1 RTC fixes:  (1) Compile issues because of missing RTC_MAGIC
        #defines, (2) missing functionality based on RTC_MAGIC in RTC based on
        stm32_rtcounter.c, (3) IRQ setup from up_rtc_initialize was later
        reset in up_irqinitialize, (4) write access to backup registers
        without enabling access to backup domain, (5) possible races in
        set/cancel alarm.  tested with STM32F103C8 only.  device now wakes up
        from forced STANDBY mode by alarm.  From Leif Jakob.
      - STM32/STM32 L4 PWM:  While attempting to output a 70 MHz square wave
        from the timer output of a STM32 clocked at 140 MHz, found that the
        reload calculation was off by one.  This correction does allow the
        output up to 70 MHz.  I am not sure this affects most users
        generating slow PWM but for frequencies close to the PCLK, the
        difference becomes significant.  From JM.
      - STM32 L4 I2C: Set I2C SDA and SCL pins to open drain mode.  From Pekka
        Ervasti.
      - STM32 L4 I2C: I2C4 was writing to wrong RCC registers.  From Juha
        Niskanen.
      - STM32 L4 DAC:  Report transfer as completed in DMA callback.  Without
        this even O_NONBLOCK writes block the calling task if DAC was using
        DMA.  From Juha Niskanen.
      - STM32 L4 COMP: comparators share RCC enable bit with SYSCFG.  From
        Juha Niskanen.
      - STM32 L4 ADC:  Correct EXTSEL macros.  From Juha Niskanen.
      - STM32 L4 TIM: TIM15,16,17 are always in APB2.  From Juha Niskanen.
      - STM32 F7 I2C: Set I2C4 SDA and SCL pins to open drain mode.  From Juha
        Niskanen.
      - STM32 F4 RTCC: ISR register and write protection fix.  From Eetu
        Nevalainen.
      - STM32 F7 Ethernet:  Fix typo in header; Add memory sync barrier
        between writing to DMA TX descriptor and restarting DMA TX. Avoid
        calling work_queue on pollwork if it's already queued, just skip a
        poll cycle instead.  Nucleo-144: Fix RMII TXD1 signal, connected to
        PB13 not to PG14.  From savinz.
      - STM32 F7:  Added missing config option for register value debugging.
        From Titus von Boxberg.
      - STM32 F7:  No FSMC, only FMC for STM32F7.  From Titus von Boxberg.
      - STM32 F7:  HEAP2 depends on CONFIG_ARCH_HAVE_HEAP2, not on particular
        FMC RAM type.  From Titus von Boxberg.
      - STM32 F7:  STM32_RCC_DCKCFGR2 has nothing to do with PLLI2S; PLLI2S is
        not dependent on LTDC, instead on SAICLK1/2 generated from PLLI2S.
        From Titus von Boxberg.
      - STM32F7: Some STM32F7 builds failed in build testing due to undefined
        STM32_SRAM1_BASE.  I think that is because stm32_allocateheap.c was
        not including chip/stm32_memorymap.h.
      - STM32 F7: dcache write-buffed mode is used (not write-through) buffer
        alignment is required for DMA transfers because a)
        arch_invalidate_dcache could lose buffered writes data and b)
        arch_flush_dcache could corrupt adjacent memory if the buffer and the
        bufflen, are not on ARMV7M_DCACHE_LINESIZE boundaries.  From David
        Sidrane.

    * STMicro STM32 Boards:

      - STM32F103-Minumum SPI: SPIDEV_WIRELESS used when this has
        changed to SPIDEV_CONTACTLESS. From Nicolas Estibals.
      - configs/:  a few more places where SPIDEV_WIRELELSS should be
        SPIDEV_CONTACTLESS.
      - STM32F103-Minumum:  Fix a BUG when reading from output pin.  We need
        a different read_ops to read from output pin. This patch fixes the
        issue.  From Alan Carvalho de Assis.
      - STM32F103-Minumum: Use separated read_ops for GPIO interrupt pins.
        From Alan Carvalho de Assis.
      - STM32F103-Minumum: Fix compiler error in MCP2415 logic.
      - STM32F746G-DISCO:  Fix for compilation of STM32F746G-DISCO.  From Ivan
        Ucherdzhiev.
      - STM32F746G-DISCO:  Fix for stm32f746g-disco board for button support
        with interrupt.  This change is tested with buttons app example and it
        is working with interrupts (signals). I tried the test with polling
        but at this point it doesn't work.  From Ivan Ucherdzhiev.

    * Build System

      - Fix -Werror=implicit-fallthrough on gcc7.  From Julien Lecoeur.
      - configs/Board.mk:  Remove comment form end of line. In windows native
        build, it appears to be trying to make that an extra parameter to the
        AR command.  From Jeff.
      - configs/Board.mk:  Remove quotes from CONFIG_ARCH_FAMILY.  Causes
        problems with Windows native build.  From Jeff.
      - Makefile.win would only create uboot images for ARM.  MIPS support
        also needed.  From Lwazi Dube.

    * C Library/Header Files:

      - C++:  Fix C++ __guard implementation for ARM.  The standard C++ ABI
        that most platforms follow defines __guard to be 64 bits.  The existing
        implementation of libxx_cxa_guard.cxx follows this.  However, the
        32-bit ARM C++ ABI defines it as 32 bits instead, and changes the
        meaning slightly so only the lowest bit is used. This matters because
        GCC creates guard symbols without regards to what libxx_cxa_guard.cxx
        says.  So on ARM, gcc allocates 4 bytes, but __cxa_guard_release
        writes 8 bytes, zeroing out another unlucky variable nearby. Fix it
        by special-casing 32-bit ARM in libxx_cxa_guard.  From Jim Paris.
      - C++:  In cwchar it uses CONFIG_LIBC_WCHAR to only export the wc/mb
        functions.  When a build does not want to use wide or multibyte char
        CONFIG_LIBC_WCHAR is not set. Therefore we should to only export the
        wc/mb functions when defined.  Regardless of the stat of
        CONFIG_LIBC_WCHAR the non mb/wc  definitions such as mbstate_t,
        wint_t, wctype_t need to be exported.  From David Sidrane.
      - C Library:  Fix an error in mkstemp() the could result in an infinite
        loop.  From Alan Carvalho de Assis.
      - C Library:  (1) Fix an error in mkstemp() the could result in an
        infinite loop.  (2)  Fix for wrong output in some cases.  For Example:
        (a) input: "FILEXX", output: "FILE00" and repeats same output for
        further invocations of mkstemp(). But, the ouput has to be FILE01,
        FILE02, ...., FILEZZ. (b) input: "FILEXXXXXX", output: "FILE100000",
        for next invocation "FILE200000" and so on.  But it's good, if the
        ouput goes like FILE000001, FILE000002, ..., FILE000101, ...  From
        Lokesh B V.
      - C Library:  gethostbyname_r: Fix check for space in buffer.
      - C Library:  inet_ntop() was printing negative values for fields >127.
      - Math Library:  Fix wrong ouput in modf() API.  The sign of integral
        part given by the modf() should be same as sign of input.  But for
        inputs between 0 and 1, the sign of integral part was not same as
        sign of input.  From Lokesh B V.
      - Math Library:  Fix wrong ouput in ceil() API.  Ex:for input x = 1.0,
        the output should be 1.0, but the ouput was 2.0.  From Lokesh B V.

    * Tools

      - tools/testbuild.sh: Fix missing $ before variable name.
      - tools/mkdeps.c:  Eliminate a warning. MAX_PATH may already be defined
        in included system files.  From Jeff.

    * Applications: apps/

      - All apps/ Makefiles:  Add .PRECIOUS: apps/libapps.a to every
        Makefile.  Hopefully this will end awkward problems when you Control-C
        out of a build and libapps.a is deleted.
      - platform/Makefile: More attempts to fix for Windows native build.
        Backslash as a delimiter causes problems in pattern subsitutions.

    * NSH: apps/nshlib:

      - NSH network initialization: Fix some ieee 802.15.4 initialization
        logic.  It should not set the saddr or panid; those cannot be
        hard-coded but must come from the coordinator.  Re-factor a function
        that has gotten too big and too complex.  Do not set the IP address
        for 6loWPAN.  The 6loWPAN stack uses IP address that derive from the
        ieee 802.15.4 addressing and cannot be (safely) configured by the user.
      - NSH Library: Fix copy-paste typo in nsh_usbconsole.c.  From Oleg
        Evseev.
      - NSH Library: fix size of 6LoWPAN extended address.
      - NSH Library: Fix build break in nsh_command.c with IPv4/IPv6 dual
        stack.  From Masayuki Ishikawa.
      - NSH Library: Correct parsing of ifconfig so that you can specify the
        HW address without specifying the IP address.

    * Examples/Tests: apps/examples:

      - examples/smart: Fix some compilation errors.  Obviously this test has
        not been used in a LONG time.
      - examples/udp:  Renaming some files to prevent name collision in
        libapps.a.  Fix naming of a configuration setting.
      - examples/nettest:  Renaming some files to prevent name collision in
        libapps.a
      - examples/nximage:  Remove unused global variable.
      - examples/nsh:  Remove APPNAME, PRIORITY, and STACKSIZE settings from
        Makefile to avoid showing nsh in Builtin Apps.  From Masayuki Ishikawa.

    * Network Utilies: apps/netutils:

      - netutils/tftpc:  Missing header file causes errors in some
        configurations.
      - netutils/tftpc:  TFTPC depends on CONFIG_NET_IPv4.
      - netutils/netlib:  Fixes to work when only USRSOCK is enabled.
      - netutils/netlib: Fix netlib_nodaddrconv() so that its return type is
        the same as other address conversion functions.
      - netutils/netlib: If only PF_IEEE802154 socket family is enabled, then
        must use SOCK_DGRAM.
      - netutils/ftpc:  Fix some memory leaks.  From Boris Astardzhiev.
      - netutils/thttpd: Fix a malformed if condition detected by GCC 6.x.x.
      - netutils/dhcpd: Prevent buffer overflow in dhcpd_addoption.  offset
        represents distance from start of option buffer.  So this should be
        changed current option pointer minus start of buffer.  From EunBong
        Song.

    * Wireless Utilies: apps/wireless:

      - wireless/ieee802154/i8sak:  Need to increment the next_saddr after
        each successful association.

    * System Unitilities (apps/system)

      - correct copy-paste typo in comments in composite_main.c  From Oleg
        Evseev.
      - system/dhcpc:  The DHCPC renew command did not build correctly due to
        naming problems.  Noted by Masayuki Ishikawa.
      - system/dhcpc: Fix warning about renew_main.o appear twice in rule.

    * Tools (apps/tools):

      - tools/mkkconfig.bat: Use %cd% instead of %~dp0 for usage of APPSDIR
        in this batch file.
      - tools/mkkconfig.bat:  Remove quotes in echo.  These were begin
        Echo'ed and generating double quots in the output.  From Jeff.
      - Windows build fixes:  tools/mkkconfig.bat - switch APPSDIR path to
        use forward slashes for kconfig-frontends. interpreters/ficl -
        Reorder some targets that causes GNUwin32 make to behave badly.  From
        Jeff.