Commit by micah :: r11940 therm/scripts/update-graphs.sh:

Atomically update the RSS feed, so there's no danger of an HTTP client pulling down a partially-written file.

Commit by micah :: r11939 therm/scripts/update-graphs.sh:

This was still having trouble defeating the picture frame's client-side cache. Added more RSS metadata to each item- now it's definitely working, yay.

Commit by micah :: r11938 therm/scripts/ (make-rss.sh update-graphs.sh):

New anti-caching approach: Generate a dynamic RSS feed with a short TTL, and include a timestamp in each image URL.

Commit by micah :: r11933 therm/scripts/ (make-rss.sh therm.rss):

Add a shell script which generates a large RSS feed which hopefully is too big for the picture frame to cache entirely.

Commit by micah :: r11932 therm/scripts/ (update-graphs.sh restart-daemon.sh):

More admin script updates- start the daemon using the correct working directory, so it can find the firmware image.

Commit by micah :: r11931 therm/scripts/ (4 files):

Use a nicer font for graphs, add a sample RSS file

Commit by micah :: r11930 therm/scripts/update-graphs.sh:

Add graphs

Commit by micah :: r11687 therm/ (2 files in 2 dirs):

New minimal rrdtool graphing script.

Commit by micah :: r11686 therm/daemon/usbdevice_fs.h:

Include a local copy of usbdevice_fs.h, since the one included in unslung is broken.

Commit by micah :: r11685 therm/daemon/receiver.c:

Include a local copy of usbdevice_fs.h, since the one included in unslung is broken.

Commit by micah :: r11684 therm/ (8 files in 3 dirs):

Replace the old mysql-based daemon with a new version which writes to flat XML and rrdtool files.

Commit by micah :: r11679 therm/ (daemon/receiver.c therm-rx.bin):

Remove some debug code, add a checked-in copy of the firmware binary to
make it easier to use this package on systems without sdcc.

Commit by micah :: r11678 therm/ (4 files in 2 dirs):

libusb sucks so much. The therm receiver has been broken for a while- I assumed it was a firmware bug, but it seems that
it was either a libusb bug or a linux bug that libusb is sensitive to. This patch ditches libusb and fixes the therm
receiver daemon.

Commit by micah :: r11096 therm/pytherm/web.py:

Add a kludge to limit worst-case memory and CPU usage. The whole architecture of this server is doomed to be inefficient, but this
hack should prevent it from completely exploding when there are a large number of samples in the database which haven't been
transferred to the RRD yet.

Commit by micah :: r10859 therm/ (4 files in 2 dirs):

More recent versions of gcc and sdcc are more strict about pointer types, and sdcc seems to handle type promotion slightly differently. These are miscellaneous fixes to the therm firmware and daemon for recent compilers.

Commit by micah :: r6514 therm/applet/thermapplet.py:

Revert that. Old therm sources don't get cleared out of the menu, and I don't feel like fixing this at the moment.

Commit by micah :: r6513 therm/applet/thermapplet.py:

Periodically refresh the sources list

Commit by micah :: r6345 therm/ (2 files in 2 dirs):

The MySQL query optimizer was scanning for all packets from a particular source then looking for the most recent ID rather than starting from the most recent IDs and looking for the requested source. This uses the FORCE INDEX syntax to hopefully correct the horrible CPU-munchingness of the server.

Commit by micah :: r6344 therm/pytherm/web.py:

Optimize out a superfluous getLatestPacket()

Commit by micah :: r6306 therm/pytherm/web.py:

Also don't blow up if a source has never received any packets