Clicking on the last build date in the tables below will take you to the associated log file so you can see what happened. Clicking on the system name will show you a history of all builds on that system.
The EMC2 tree is the focus of new development. The head of the tree is the mainstream version of EMC2. Other branches are used for experimental work, things that will break the main tree, or variations that need to remain independent of the main tree for one reason or another.
The v2_2_branch is the stable release branch for version 2.2. At this time it is only getting bugfixes, as we prepare for the release of version 2.3.0.
Added another 2G of ram to the farm host, and added four new virtual machines. The existing Dapper Realtime system has been moved from a real box (that has been powered down a lot lately) to one of the VMs. The other three are now running non-realtime Dapper, non-realtime Hardy, and realtime Hardy.
8-Nov-2008Removed the EMC-2.0 and EMC-2.1 branches. The last commit to either of these branches was nearly a year ago, and we don't anticipate any future commits.
11-Nov-2007Added the EMC-2.2 branch. I should probably remove EMC-2.0 as well. Maybe later.
17-Jun-2007The farm is faster now. I've moved the virtual machines to my new computer, a Core2 Duo with 2G of RAM. The old computer with only 512M would get into a serious thrash-fest if multiple builds were going on at once, sometimes taking up to 45 minutes to finish. The new one takes 2-3 minutes max. Mostly thanks to plenty of RAM, but I'm sure the CPU helps.
3-Dec-2006Turned off slots 3 thru 5. My basement is really quiet now. According to the logs, each of those slots compiled CVS head about 1200 times over the last 13 months or so.
2-Dec-2006The rest of the new compile farm slots are online.
Slot 6 is a VMware virtual machine running BDI-4.51. It requires non-standard build scripts, because of various weirdness in the BDI kernel and system configuration (configure needs help to find RTAI).
Slot 7 is not a virtual machine. It is a dedicated user account on my main Ubuntu 6.06 LTS (dapper) machine (which has a realtime kernel).
Slots 3, 4, and 5 are living on borrowed time - I will be shutting them down one of these days.
25-Nov-2006The version 2.1 branch has been created, and the new compile farm slots are now building it.
25-Nov-2006The first two virtual compile farm slots are online. They are VMWare virtual machines, running on my main PC.
Slot 1 is running Ubuntu 5.10 (breezy), without any realtime patches. It only builds CVS HEAD, because version 2.0.x doesn't support non-realtime simulation.
Slot 2 is also running Ubuntu Breezy, but it has the realtime kernel and is building a complete EMC, both HEAD and the version 2.0 branch.
17-Nov-2006Slot 2 died... the ancient 2G SCSI disk started losing blocks. We're only a few weeks from moving to a new farm using virtual machines, and we have also decided to stop supporting the very old systems, so I'm not going to attempt to rebuild that slot.
The log file shows 1104 builds of EMC2 on slot 2 since November 2005. (Thats when I started logging historical results.) The farm has actually been running since April 2004, and it also built EMC1, RCSLIB, and the EMC2.0 branch, so it probably has several thousand builds under its belt. Not bad for a system that was already 5 or 6 years old when I literally dragged it out of a trash box.
27-Oct-2006"Slot 7" wasn't really a slot on the rack, it was running on my main PC. I just upgraded to Dapper, so this slot is no longer running. I am planning to move the entire farm to virtual machines running on the main PC over the next month or two. The plan is to have slots for Ubuntu Dapper and Breezy, and BDI-4.xx. I also want to have at least one slot building the simulator version on a system with no realtime. The older systems (BDI-2, BDI-TNG, and BDI-Live) will no longer be supported.
17-Apr-2006
Major revisions to the farm scripts. In addition to sending the results to the webpage, they now send failure notices to the #emc IRC channel, and to the emc-commit mailing list. The old approach wasn't very effective - sometimes a problem wouldn't be noticed because people weren't regularly checking the webpage, and it would become a major task to go back and figure out exactly what change broke the build.
The new scripts are also a lot more modular, and I finally got around to writing a README that covers the steps needed to set up a farm slot from scratch. The farm scripts are also now stored in a directory on our nice new CVS server.
Removed EMC and RCSLIB from the farm. EMC(1) and RCSLIB don't work on 2.6 kernels because of the new kbuild system and other extensive changes. We're not doing any new work on EMC(1) anyway, and we're certainly not going to spend the time and effort it would take to get it running on 2.6 systems.
Out with the old, in with the new.... along with new scripts, we now have a new target. The farm is now regularly building the v2_0_branch, which will eventually be released as version 2.0.0 of EMC.
22-Nov-2005I modified the farm scripts to keep a historical record of the pass/fail results from each build. A single line containing the date, time, and result, is appended to a file after each build, and the file is accessible from the webpage.
The EMC compile farm is maintained by John Kasunich