* Add embARC EMSK BSP in third_party folder.
* Add synopsys in Makefile.am for embARC BSP
* Add platform implementation on ARC EMSK
* Add EMSK in Makefile.am and Makefile-emsk
* Add EMSK support in apps
* Modify configure.ac to add ARC EMSK.
* Add support for the Host Power State feature.
* Fix compiler ewrrors.
* Fix compiler errors round 2.
* Add newlines to end of file.
* Address review comments.
* Arrange to send the power state response again if the first attempt fails.
* Fix compiler issue.
* Add otPlatRadioGetReceiveSensitivity() API
* Add a new otPlatRadioGetReceiveSensitivity() API to get sensitivity value;
* Use the receive sensitivity value as the noise floor for link metric computation;
* Remove some unused noise floor related functions.
* Add SPINEL_PROP_PHY_RX_SENSITIVITY Spinel/NCP property
Enable OPENTHREAD_CONFIG_ENABLE_STEERING_DATA_SET_OOB to be able to set
steering data out of band on FFDs.
On joining device, enable filtering MLE discovery responses based on Factory assigned EUI64.
* Unify the location of "#if OPENTHREAD_FTD" which include/exclude the whole file
Also correct some comments in xxx_api.cpp
* Remove Address Resolver related features from MTD build
* Remove some Leader/Router related features from MTD build
* Remove Leader/Router related APIs from MTD build
* Remove Leader/Router related functions from MTD build
* add pskc api
* add pskc cli
* add NCP api of PSKc
* apply PSKc from network
* implement PSKC api for windows
* add name for pskc property
* clear active and pending dataset when settings pskc
This commit contains node logic change. It makes some formatting
changes in the comments, also renames method/function argument/param
variable names to follow the style guideline (e.g.,
`anInputBuf -> aInputBuf`).
* PR-1599 - examples/platforms/posix/flash.c fixes
* Some notes:
a) Do not change this to a for(;;) loop that writes bytes
This takes forever and causes timeouts to occur.
The math:
With pagesize = 8K pages, and n-pages=128
This turns into 1 million calls to "pwrite()"
Those 1million calls occur during startup.
the "Thread Cert" tests, in some cases start
over 30+ instances of the simulated app...
Tests actually run in parallel, in some cases there are
70 to 80 instances of the CLI app starting up. Each one
calling write - 1 million times, writing exactly 1 byte
That becomes super slow. Thus we write full pages
On a powerful platform this is fast, very fast.
On a virtual machine (ie: test-build-farm) it is slow
Result is: The Thread Cert tests fail/timeout
b) Do not allocate memory for the page via malloc/free.
Reason: this causes false failures in some test modes
Specifically when enabling the address sanitizer ...
And running thread certification tests *in*parallel*
The memory requirements become huge.
Background:
The way "asan" works is this: When you free a page
the "libasan" does not actually free the page instead
libasan poisons the page but does not actually free the
page, why? Because the goal to detect using the memory
Thus ASAN is like a giant memory leak, it must hold onto
the page so that it can detect "use-after-free"
Next (part 2)... the CERT tests, and run in parallel there
are cases where 70 to 80 instances of the CLI app run
in parallel - its a huge pseudo-memory leak times 70
Depending upon your machine you need 8 to 16gig of ram
Virtual test boxes are often configured with much less
The solution is to use a full page on the stack
it could be a static, variable however, we are
running on Linux, we have plenty of RAM..
- Invalidate Active and Pending Datasets when configs are changed via APIs.
- Do not allow client to change network configurations while Thread is enabled.
Create a new specialized ChildInfo structure, which will be only used for children information storage
Also update otPlatSettingsGet() to be consistent with its documentation:
* return the actual length of the setting, but not the actual length it read to aValue
This commit changes build behaviour, so that platform-utils library
is always built. This allows for using opentread in environments which
keep platform files out of openthread tree.
* Fixes some Spinel usage in otLwf and some logging in the core library.
* Undo cc2538 config logging change
* Fix build error for nrf52840 with all logging enabled.
* Update minor version check based off feedback
This commit includes two changes:
* Removes existing `AC_NO_EXECUTABLES` from `configure.ac`, since it
is a hack only intended to be used by the GCC project.
* Adds a new configure argument: `--enable-no-executables-hack`,
which restores `AC_NO_EXECUTABLES` for broken build environemnts
which require it.
* Removes `--target` from example makefiles, since it is only
appropriate when compiling compilers.
As a side effect of the second change, cross compiling with the
makefiles now no longer causes the `arm-none-eabi-` prefix to be
prepended to all built executables. This change updates all of the
documentation accordingly.
* Fix otSendMacDataRequest API
* Add otSendMacDataRequest declaration to API headers
* Move otSendMacDataRequest from thread to link API
* Update otSendMacDataRequest to otLinkSendDataRequest
Jonathan Hui