Last Monday, I began a new project at Codethink - working on kernel support for the next generation of an interactive system fitted in some high-performance cars. I travelled up to Manchester the night before, and on Monday I met Codethink's team (most of whom I was familiar with).

My first chore was to set up VPN access to the development resources (source control, wiki, etc.). I sandboxed the proprietary VPN client in a VM with a systemd unit to run it at boot, so I can control it by starting and stopping that VM. I then set to work on unpacking and exploring the SoC vendor's evaluation module (EVM), starting by looking at serial output - of which there was none. Nothing on the LCD panel or network port either. A frustrating day.

On Tuesday, the architect from our immediate customer, already a friend of mine, joined us in Manchester. As for the EVM, without any obvious change in cabling it was now reliably spitting out the expected u-boot and kernel boot messages, unfortunately ending with errors from the touchscreen driver intermingled with messages from various graphics drivers. I looked at the u-boot environment to see how the kernel parameters were controlled and started tweaking them with the aim of working around the errors. None of this worked. I looked for more documentation on the EVM, but found none beyond the quick start guide. Had I been given the wrong kernel or device tree? They matched the available downloads though the device tree looked very different from the upstream version. I started trying to edit the vendor's device tree to remove the troublesome devices before realising that the decompiled tree was full of absolute cross-references that would be invalid after recompiling. So, a second day of frustration. In the evening I started to feel a cold coming on.

I slept poorly, and just dragged myself out of bed in time for 11:00 checkout. In the office, I remembered that u-boot had commands for manipulating device trees in memory. I tried stepping through the boot script and then removing device tree nodes for the touchscreen and graphics hardware before booting the kernel. Success! I got a login prompt where of course username:root, password:[empty] worked. (It doesn't run sshd, but does run telnetd.) Then it was time for a team lunch at the excellent Samsi, though I hesitated to go as I was feeling quite poorly.

Shortly after lunch I found myself falling asleep at my desk in the middle of reading documentation. I had to take a nap on the sofa. I got a little more work done that day, but I got progressively feebler. The train journey home was thankfully uneventful. I arrived home at midnight and soon went to sleep.

I spent most of Thursday in a high fever, dreaming of abstract entities and their relationships that I struggled to name, but in retrospect might correspond to various kinds of software component. It was incredibly tedious. I had no appetite and little ability to process the real world. Nattie tended to me as she could. I called my mum to cancel her visit planned for the next day.

On Friday my fever started to recede and I drank some soup. I was able to look after myself to some extent, though I sorely missed Nattie when she went out. I spent most of the day watching comedy and music docs on the BBC iPlayer, something far too passive for me to do normally.

On Saturday morning I felt well enough to tackle the meze platter that Nattie brought back from her previous evening's outing. It tasted pretty damn good. I felt better hour by hour, and in the evening we both went out to a group dinner in a pub. Hot food smelled and tasted even more amazing! I was able to start catching up on email and some real work, preparing kernel stable updates. I even migrated DNS service for decadent.org.uk late that night, driven by anger at the incompetence of my previous provider that had prevented my mails reaching bugs.debian.org.

Despite the awfulness of Thursday and the symptoms I still have (tickly throat and runny nose), this bout of the 'flu has been a positive experience for me, but I'll explain why that is in a subsequent entry.