veyepar: Life in the Armoury

Author(s):	Peter Chubb
Location	Wellington Room 1
Date	jan Wed 18	Days Raw Files
Start	22:40	First Raw Start	22:32
Duration	0:45:0	Offset	0:07:35
End	23:25	Last Raw End	23:32
Chapters	00:00 0:21:54
Total cuts_time	49 min.

http://linux.conf.au/schedule/presentation/170/
raw-playlist raw-mp4-playlist encoded-files-playlist host archive tweet mp4 svg png

assets release.pdf Life_in_the_Armoury.json
logs
Admin: episode episode list cut list raw files day marks day marks day image_files

State:

Locked:

clear this to unlock

Locked by:

user/process that locked.

Start:

initially scheduled time from master, adjusted to match reality

Duration:

length in hh:mm:ss

Name:

Video Title (shows in video search results)

Emails:

email(s) of the presenter(s)

Released:

has someone authorised pubication

Normalise:

Channelcopy:

m=mono, 01=copy left to right, 10=right to left, 00=ignore.

Thumbnail:

filename.png

Description:

markdown

Our group does seL4, eChronos and Linux kernel development across many different embedded  platforms, including various ARM, aarch64, PPC, and Intel.
Even though the core of the released seL4 kernel is formally verified, the experimental versions are not, and need to be tested on all these platforms.  In addition only one of the platforms corresponds to the verified kernel.
We have a very limited budget, so often have only one of a platform.
This talk explains how we do continuous integration and continued development without going mad,
We have developed custom hardware (the switcheroo) that is essentially an arduino-controlled multi-output high current power supply, use commercial-off-the-shelf telnet to serial gadgets, use a Cubietruck running Debian to boot some things via fastboot, but everything we can we boot via TFTP.
Clever python scripts talk to everything as glue.
Interesting problems include hard-wired fastboot identifiers (what do you do when three of your development boards identify as 123456789abcdef as a fastboot ID?), multiplexing between different users (the obvious solutions have problems, and we still don't have a good way to do this), and the need for custom u-boot that means that sometimes the things we test don't match what's available to the average punter.

Comment:

production notes