CVS log for emstar/timesync/BUILD
	Help

Request diff between arbitrary revisions / Display revisions graphically / Jump to LXR

fixed bug in gsync caused by parameter timeout

fixed busywait to be efficient for longer waits
libroute uses new timesync api call
added gettime attime to copydeps for timesync

added two utility programs and added sync_clock_no_create() api call

Lots of minor fixes to get ready for deployment

added sync parameter pruning..

minor fixes to make shlibs compile

Big commit -- the bulk of the Cygwin compatibility is being checked in here.
The Cygwin port doesn't really run yet since FUSD hasn't been ported, but,
with this checkin, the rest of the repository compiles.

fixes to get this to build .so's correctly...
patch from dustin

take old syncview out again

get timesync to work with emview

added status device for lessgps

renamed output driver source files

* Finally fixed timesync pairs API; now it's a device you open and write
pairs info to

 * Changed audio server to use new pairs API

 * Cleaned up and sorted out the tangled mess that was the include files
 for timesync; all userspace programs should only include sync.h now, and
 the daemon should only include syncd_i.h.

 * Moved some time conversion routines from sync header files to misc_time.h

 * checked in pair-insert.c, a simple program demonstrating the use of the
 new pairs API

 * Changed pulseat.c to use misc_time functions rather than the old circlemud
 functions

Merged many timesync fixes in from the lessgps_release branch

add logdiff program

added rbs time driver to lessgps

Extremely simplified sensor-client interface. Now, requesting
samples is a one-line request that returns a simple structure.
Please see sensor_client.h for more details.
sensor_dev_client.c reflects these changes, and demonstrates
many of the possible usages.

Made an NMEA output driver for LessGPS

build lessgps

* fixed problem with build files when ezbuild semantics changed for copy blocks
* added std.run
* added support for additional binary dependencies that will work when
  the variable stuff is fixed

moved to new non-broken paths.
fixed  hostmote so that it remembers settings and can correct them
if the mote resets

build pulseat ONLY for stargate

build pulseat

Added a test-feature to sensor-client. I am in the process of
creating a stand-alone test that will test ring-buff, sensor-device
and the timehist buffer.

Fixed  bugs in sensor_dev
- where if the request
was in the past in some cases it wasn't getting satisfied.
- sample number 0 was always getting skipped

Took out unused code, and merged files in:
sensor_dev_util.c + sensor_dev.c => sensor_dev.c

sensors/wmr/ - now only has:
wind_client.c, wind_dev.c, wmr_dev_util.c -
wmr_i.h

Changed structure of sensor-device a bit - it now retreives data
and timing information before the data-call-back. This can
be more efficient, but for now it works more accurately for
certain cases.

Added client functions to timehist so that there is a funtion
interface to retrieving data, closing timehist buffer.

Added a helper function to sensor_client that prints out the
data for the binary call-back. This was just added for debug
purposes because the binary interface keeps changing and
it is causing confusion. (sc_print_bin_output)

Added a catch in sensor-device so that - even if the timehist
daemon is not running - as long as requests are made in terms
of sample-numbers - then it won't fail. And if the request
is made in terms of time and the timehist daemon isn't running,
then it won't crash.

Fixed a bug in ring-buff where it woudn't return the correct
number of samples in certain instances

Added local_libs to BUILD files in libdev and sensors to prevent
linking errors in query_client, when linking sensor_client and
wind_client.

Took out all time related functionality from ring-buffer and
moved it to a new timehist device service. Any client can
access a device if they know which device is
associated with the sample-clock they are interested in.
/dev/timehist/service handles the opening of new devices - which are
then created as: /dev/timehist/service/<sample_clock_ID>

Future Work:
- Remove limiations on the number of timehist-devices that can be
created
- Make the data stored in timehist-device stored in a ring-buffer.

MINOR:
- Completely restructured sensor_dev code because it was becoming
too unmanagable. In the process I cleaned up small issues/bugs. Too
many to detail here.
- Minimized the usage information that is printed out when a sensor-device
  is catted. Now, to get detailed information use: echocat <device_name> usage

	- Fixed sdev_example so it truly took the sample size from the command
	line.
	- Fixed spacing in all of my files. Made them look much prettier.
	- Formalized the interfaces between ring-buffer and sensor-device and
	timehist and sensor-device.

removed deprecated visualizers from the build.
remvoed dependencies on gtk+1.2

Build res-test kernel module

Cleaned up syncd's argument processing; converted from getopt over to
Lew's misc_parse_out_option stuff which really is superior now that I
bothered to learn how to use it

syncd can now be told to use a particular network interface (e.g., eth0)
when using the UDP driver.  Although, I think the UDP driver will be
deprecated in favor of running over UDPd.

Added a simple program to determine the clock resolution of the kernel

syncd now maintains a dynamic directory of registered clocks.  It uses
the new 'directory device' service in libdev.  'Mote' and 'codec' types
have been removed; drivers that provide pairs information are now expected
to register their own clock-types with the timesync daemon using the
directory API.

The MoteNIC has been changed in this way... it now registers a clock that
has the name of the mote's data device (e.g. /dev/mote/0/hostmote).

fixed generated filename

gsync milestone -- first draft, not tested yet

link interface for timesync -- timesync now is integrated with emstar,
and works in the simulator!  hooray!

Factored beacon-sending code out of driver_udp and into a new file
syncd_server.c, so that the new emstar link driver can be added without
code replication

libpcap used in syncd is now deprecated; we're using SO_TIMESTAMP instead

build logic analyzer control module for experiments

Did a lot of cleanup of the initialization code.  Took out e-run and
replaced with misc_init.  Updated some older programs to use emrun_init.

added -lm, needed when compiling without -O

Added easybuild features:
   -- platform-specific targets now supported ("build bins for i686-linux")
   -- protection against specifying redundant target names
   -- easier to specify platform-specific global cflags/ldflags
   -- other minor fixes

timesync doesn't quite "dual-compile" yet, still one missing feature
(multiple .o compliations with different cflags) which will be added
shortly.

link with -lm needed for arm

more fixes...... does this actually WORK???????

many more changes; hopefully got a handle on the recursive dep problem
between .d files and the directories that they sit in.

started the process of converting the code itself to conform to the new
build system (mainly fully-qualifying include filenames)

more checkpointing -- relative dirs work, the basic structure of bootstrapping
everything from the root directory works, more black magic with gcc -MM
stuff, objdir tokens integrated, concept of 'native build' on which the
easybuild system itself runs.... lots and lots of details.  but then
again, is a build system really anything BUT lots and lots of details?

add prelim build specs