Oct 10: We discovered that we can't always just copy programs from the device to the emulator. Sometimes packages rely on other libraries. We figured out a way to remount the primary yaffs2 partition of the emulator as RW, allowing us to copy over library files. Unfortunately, this hasn't resolved our issue.

Oct 12: Requested MySQL database from CSL. More work on client that sends proc data to server. Started looking into how to schedule background processes on the phones.

Oct 15: CSL setup the MySQL database, and we got a basic database started. We setup a php page on our CS space to get POSTs from Androids and do the inserts, and the Android application can send data to the page.

Oct 27: We figured out how to create Android services. We also figured out how to make an application receive a message that is broadcast upon startup so that the application can start our service. Unfortunately, although the application works, it apparently hangs for a while, as we get a “force quit” or “wait” message upon startup. Clicking “wait” allows the application to work. Update: The application not responding message appears to not occur on the actual device.

Oct 29: We significantly modified our two logging programs/services (one records data and the other uploads data to our server). The redesign design is simpler, will fix a concurrency issue, and should be faster, but it's not completely working yet.

We added more tables to our database where we will dump more procfs data.

We looked into getting more interesting information from Android in addition to basic procfs stuff and found a BatteryStatsService class (frameworks/base/services/java/com/android/server/am/BatteryStatsService.java) that could potentially be extremely useful by providing battery usage information.

We did more work on our program that automatically runs a bunch of applications on the Android, but it's still buggy. We observed that some of the applications we were trying to start were already running when the experiments began (they begin at startup), so killing those first helped. We plotted the memory usage of some of the applications we were able to collect data for and noticed some interesting trend. Primarily, VM usage far exceeds what is ever kept in memory. We have yet to explain this.

Oct 30: We met to further troubleshoot our logging/posting services, but haven't made further progress.

Nov 5: We finally found and fixed several bugs in our code (mostly formatting issues) that had been causing problems for quite some time. We also combined our logging and posting programs into a single program. This makes the code simpler and eliminates the need to worry about concurrency issues. We discovered that some of the issues we are still having are due to Android emulator procfs content being slightly different than the formatting on actual devices.

Nov 6: We finally figured out how to properly create services that don't result in the “force quit”/“wait” dialog. onCreate needs to return quickly instead of looping indefinitely, so we set up a timer to cause work to be done at regular intervals instead of doing everything in onCreate.