klueska [Thu, 15 May 2008 23:57:13 +0000 (23:57 +0000)]
Reverting back to decentralized handling of interrupts on the msp430. Centralizing them has the side effect of forcing all interrupt handlers to be included regardless of whether the component that actually DEPENDS on it is included or not. Leads to unnecessary code bloat. Original motivation for centralizing them in the first place was in anticipation of adding tosthreads in the upcoming release. A better way of supporting them without the need for centralized interrupt handlers was found, hence the rollback.
rincon [Wed, 14 May 2008 21:33:07 +0000 (21:33 +0000)]
* Removed hardware address recognition to monitor its affect on unit tests, low power listening reliability, and ack reliability in general.
* Implemented software-based address recognition in ReceiveP.
* Applied Steve Dawson-Haggerty's patch to make the RadioBackoff interface parameterized by AM ID only at the top of the stack, whiich allows stack modifications that do not support AM ID's to exist below the active message layer.
andreaskoepke [Tue, 6 May 2008 14:19:36 +0000 (14:19 +0000)]
incorporate a patch from Frederik Hermans -- looks like the 2.4 kernel can not
send all 260 bytes of a normal msp430 program chunk at once, so now you can
choose a proper chunk size on the command line.
Allow possibly multiple slaves for the Atmega128 SPI bus.
Move the slave select logic from the HPL SPI to the user of the resouce.
This touched the MicaZ SPI access (but not the mica2 which uses its own logic)
Added default command so that the entire parameterized AMSend interface doesnt have to be wired to when instantiating this component. Useful for reusing this component to build a second layer of queuing on top of a single AM type
instead of exiting, print just a warning when data section is not found in the binary (e.g. Blink compiled with avr-gcc 4 will not contain a .data section)
bugfix: the buffer passed to ReadStream.postBuffer() must have a certain minimum size to store temporary information (this commit happens in agreement with David G.)
- bugfix: ReadStream did not work with a "usPeriod" parameter > 0xFFFF (the fix uses/adapts the atmel ReadStream code)
- bugfix: overflow interrupts were not handled/signalled correctly from the HAL
- fixed the wiring for DMA, which was not correct when used by more than one client
- polished the HPL and added some comments
- introduced a compile time warning when DMA/TimerA is accessed (will be removed once we decide how these components are arbitrated)
klueska [Thu, 27 Mar 2008 00:52:02 +0000 (00:52 +0000)]
Update to allow header files to be processed via standard gcc and not have to run through ncc first (problem was in the usage of @combine, etc. in these files)
klueska [Tue, 11 Mar 2008 09:51:22 +0000 (09:51 +0000)]
Changed bsl.exra to search for the string between the commas and not short circuiting after seeing a digit when installing to some default tty device. Also added some more variables to the avr.rules and msp430.rules to allow more flexibility in the way the tos-storage-XXX scripts are invoked
klueska [Tue, 11 Mar 2008 09:34:13 +0000 (09:34 +0000)]
Changed bsl.exra to search for the string between the commas and not short circuiting after seeing a digit when installing to some default tty device. Also added some more variables to the avr.rules and msp430.rules to allow more flexibility in the way the tos-storage-XXX scripts are invoked
klueska [Tue, 11 Mar 2008 08:28:54 +0000 (08:28 +0000)]
Changed bsl.exra to search for the string between the commas and not short circuiting after seeing a digit when installing to some default tty device. Also added some more variables to the avr.rules and msp430.rules to allow more flexibility in the way the tos-storage-XXX scripts are invoked
janhauer [Wed, 5 Mar 2008 18:40:17 +0000 (18:40 +0000)]
Bugfix for a bug in the timer system that was first reported here:
http://sourceforge.net/tracker/index.php?func=detail&aid=1572439&group_id=28656&atid=393934
It showed up when you used two Timer interfaces (e.g. provided by two TimerMilliC() instances) and Timer2.startOneShot() was called withinTimer1.fired() and in addition (1) Timer2 had a higher internal interface ID (unique(UQ_TIMER_MILLI)) than Timer1 and (2) there was some (small) amount of time elapsed before the Timer2.startOneShot() was called in the fired eventhandler. Then, Timer2.fired() was signalled immediately independent of the "dt" parameter in startOneShot(uint32_t dt). This has been fixed now.
For more information see:
http://mail.millennium.berkeley.edu/pipermail/tinyos-devel/2008-March/002573.html