idgay [Wed, 6 Jun 2007 22:40:17 +0000 (22:40 +0000)]
build both 32 and 64-bit versions of the JNI code so as to support both
32 and 64-bit JDKs
install the appropriate JNI version depending on whether the JDK is 32 or 64-bit
rincon [Thu, 24 May 2007 19:55:12 +0000 (19:55 +0000)]
Previous problem: The EventDispatcher is stuck on the function waitForEvent() while we are attempting to close. Closing NativeSerial while the driver is still in the function waitForEvent() will crash everything. Calling cancelWait() in win32 does let waitForEvent() return until some event actually occurs, which is not the behavior I was expecting. The serial driver itself could be fixed to let cancelWait() actually force waitForEvent() to return false, but I don't want to touch it.
Solution: Force waitForEvent() to generate an event by adding a notification for OUTPUT_EMPTY. When we want to close the serial source, send a 0x7E byte to the serial port and wait for that OUTPUT_EMPTY event to get signaled. At that point, the EventDispatcher thread can continue execution and will cancelWait() properly. We aren't stuck in waitForEvent(), and only then can we shut down NativeSerial with confidence.
One other issue remaining is the fact that sometimes you can't reconnect to the source very quickly after a disconnect. Adding in a hacky wait(500) fixes the problem on disconnect, but I haven't included that anywhere because I haven't figured out exactly what is causing us to need to wait for a short period of time after a disconnect before reconnecting again.
razvanm [Tue, 22 May 2007 20:34:19 +0000 (20:34 +0000)]
Initial commit of the Deluge T2. Some notes:
- TOSBoot includes some code to allow the MicaZ version to compile but it's not support for it it's not done yet.
- tools/tinyos/misc/Makefile.am is not updated yet so the new tools will not be installed by default.
scipio [Mon, 21 May 2007 21:35:54 +0000 (21:35 +0000)]
Improved accuracy of acknowledgement delivery by considering SNR
curves. Currently, ack reception is based on the same SNR/PRR curve
as packet delivery. This is pessimistic, due to the fact that acks
are shorter than packets. I need to measure the ARR/SNR curve to
put it something more accurate. Also, currently TOSSIM does not model
false positives, something I will be adding shortly.
rincon [Fri, 18 May 2007 18:27:03 +0000 (18:27 +0000)]
Previously, the serial forwarder would only release the COM port on exit. Now it provides the ability to temporarily release or reconnect to the COM port when you close/open your source. The SF GUI demonstrates this ability when you connect to your mote, stop the server and verify you can reprogram your mote at that time, restart the sf server and verify you can continue communicating with that mote... all without closing down your serial forwarder.
Fixed SPI bus crashing issue by adding in a SplitControl interface to shut it down before turning off the radio. The majority of this work and debugging was spawned by Roman Lim.
- instrumented for performance evaluation
- changed SleepTime interface to allow a local sleep time that is smaller
or equal to the largest sleep time in the network, ideally there will be
a time when Sleeptime can be replaced with LowPowerListening, but currently
the semantic of LPL does not fit
- changed initialization of variables
Allows proper generation of tos_image.xml.
- fixes a small mistake in support/make/Makedefaults that
prevents the generation on ident information
- updates support/make/tos_image.extra and tools/tinyos/misc/tos-write-image.in
to allow generation of the platform information in the tos_image.xml.
Fixed TOSSIM bug where you have to reference this Atm128Timer.h in
funny places due to component structure differences. JUst include it
where you should, that is, where it's used.
Update of Std/Split Control interface definitions to be in line with TEP115. Also added the EALREADY type to TinyError.h so that providers of the SplitControl interface can return this value