Contributed by Paul 'WEiRD' de Weerd on from the cheese-induced-hallucinations dept.
Fresh from Bucharest is this story from Martin Pieuchot (
mpi@) with his experience from p2k19:
Since I attend OpenBSD hackathons, I hear stories about how crazy are the ports hackathons. So I try my best to look like a porter in order to experience this craziness. I must admit p2k19 was awesome but the craziness of port hackathons is still an enigma to me.
I arrived after a pre-hackathon at Miod's, also known as the french cheese challenge, where I shared my plan with ajacoutot@: provide debug info for all the ports.
It is obvious that complex ports are the pieces of software exposing many bugs and limitations in OpenBSD. Providing debug information for these ports will result in easier debugging which should lead to better software.
So my plan for p2k19 was to discuss the matter with porters to see if and how this could happen. It turned out everybody talked about the subject at the first dinner, even if we didn't eat altogether ! You can imagine the smile on my face on the second day when I received the first implementation in my mailbox :o) Thank you!
I spent most of my coding time working on debugging and tracing tools on OpenBSD. From profiling to ddb(4) while adding features to a tool similar to Dtrace and BPFtrace.
I profiled an interesting scenario where doxygen(1) ended up make 4 CPU spin on the KERNEL_LOCK()
This tool is already useful to visualize contention, like in the graph below where a 4 CPU machine is building a libc
During the week I worked at adding some missing features to debug the race exposed by recent UVM changes.
I also got the chance to present one of my debugging stories and talk about locking to students of the Operating System course. This was a nice experience.
Overall this has been an awesome hackathon, thanks for the great weather, thanks for all the food and drinks, thanks for the positive vibrations, the guitar, the nice place, the hosting. I'd love to come back :)
(Comments are closed)