Using clang (on osx) yields to some warnings
(conversion between signed/unsigned, unused variable)
which gcc would only produce when used with -Wall.
This commit fixes those.
When running multiple tunslip6 instances, it collides with tun0 being already
used (cannot open file).
However, system default is already to use "tun0" and "tap0".
By putting the default empty string as name, system automatically increment and
selects a free interface, "tun0", "tun1".
Previously, the Cooja mote assumed that its file was always initially empty (file.endptr == 0). Therefore, a file uploaded to a mote's CFS could never be read by the mote, as the mote would prevent reads from going past the EOF (indicated by endptr).
By tracking the file size and making it accessible to Cooja, the correct size of the uploaded file can be reported to the mote and allow it to read the uploaded file.
Refactored RadioMediumObservable to RadioTransmissionObservable because of its function
Added correct RadioMediumObservable and updating in AbstractRadioMedium, DirectedGraphMedium and MRM
Added some documentation
The .2mg image format contains a header which is missing from our file. So our file is rather a .po image.
I opted to not add the .2mg header as it is only necessary if the metadata it contains differs from the values "guessed" when using the "naked" .po image format. On the other hand there are image file consumers not understanding the .2mg image format.
- By end of Jan 2014 SuperTweet.net was shut down (http://supertweetnews.140plus.com/). So Breadbox64 has come to an end for sure :-(
- The email app - or rather email sending app as it is SMTP only - can't be used anymore since nowadays everybody uses some "strong" authentication for SMTP session logon (thanks NSA).
- The ftp client app isn't very useful as it supports only download - for which the WGET app is almost always more useful for. But more important it doesn't support PASV which is more or less the only supported mode nowadays (especially over NAT).
This patch restores the original behaviour of Cooja when
the transmitted packet is correct (which is true in a vast majority
of cases).
In case of a wrong outgoing packet (wrong length, wrong preamble)
the transmission will end when the radio changes its state (which
should always happen after transmitting a packet).
Benchmarks with RPL (33 runs, 50 nodes, 3 hours of simulated time each)
yield the same results (PDR, delay, number of transmitted packets)
as with the unmodified Cooja.
The packet converter used to generate packets of length zero
when it encountered errors during conversion. This caused
exceptions in packet analyzers.
Now the converter returns null in case of error. Appropriate
checks have been added to the code that uses the return value.
Due to errors in mspsim and/or radio drivers, packets of incorrect
length are sometimes transmitted. The length might be larger than
the 127-byte maximum (considered negative in the current code)
or not matching the actual number of transmitted bytes.
This leads to wrong packet delimiting when converting from
the mspsim-level stream of bytes to Cooja-level packets
causing unhandled exceptions that terminate the simulation.
This patch checks the frame preamble (0000007A) and the length field.
If they are wrong, no decoding attempt is done.
The transmitted bytes are still delivered to the receivers untouched.
The connection is terminated when the radio state is changed (which alway
s happens when TX is done).
* Fix CCA detection in Cooja in the case when the receiver swicthes on the right channel during an ongoing transmission. Always add a connection on transmission, even when the channel is not correct. Initially the connection is in a dormant state; this mimics what Cooja is doing when the receiver radio is turned off;
when the receiver is turned on and switched to the right channel, `updateSignalStrengths()` is called, and the connection starts to recieve PHY-level traffic.
* Add "channel" property for DGRM edges.
* Avoid cross-channel interference on DGRM and MRM radio mediums
In the previous implementation a click event was triggered when the
button was pressed.
This implementation allows to set and release buttons independently both
by mouse clicking and by key typing.
SIM_CONTROL_PLGUIN are handled like SIM_PLUGIN, with one exception.
If the simulation is started with -nogui=<simfile> than it is checked
if a controlling plugin is loaded. If not, the simulation terminates
with an error message.
The implementation of clickButton(), pressButton(), and releaseButton()
can be shared accross the several node-dependent implementations as
they use the node-dependent doPressButton() doReleaseButton() routines.
As every Button has a node-specific implementation part, this should be
the minimal interface to the backend node emulator for pressing and
releasing a button.
Previously getOutputPowerIndicatorMax() returned the fixed value 31.
This is valid for e.g. the mspsim CC2420 radio implementation but not for the CC2520 implementation where the maximum returned value is 9.
Thus to fix transmission range issues (for example for Wismote node) the maxium value provided by the radio implementaiton must be used.
