Contributed by Nayden Markatchev on from the gnome galore dept.
Next up with a p2k18 report is Jasper Lievisse Adriaanse (jasper@):
When the strikes were announced I decided to change my travel plans a bit and ended up arriving in Nantes one day early. That made me avoid the strikes that were on strike, oh well.
After having sampled local brews the night before with tb@, the hackathon was already underway when I showed up in the hackroom the next morning. With GNOME 3.28 having been released a short while ago, it was clear what ajacoutot@ and I would be in for...first finish your veggies, then you can have desert; first update GNOME, then you can hack on other things.
In the past few years we've gotten quite used to this and as such we managed to get nearly all GNOME 3.28 components committed before the end of the second day. Some runtime fallout was expected, which we'd deal with once the tree was fully building again and we'd have fresh packages. One particular issue turned out to be both buildtime and runtime breakage, and with the help of the terrific pudb debugger for Python I was able to figure out the cause and solve it in gobject-introspection.
By the end of the hackathon I'd find myself back in gobject-introspection again to debug another issue with g-ir-scanner. Turns out that parsing ldd(1) output with Python across multiple platforms and for multiple build systems has its quirks. Who knew?
Aside from assorted cleanups (dropping ancient GNOME2 libraries, cleaning up openbsd-wip, chasing breakages) and updates (Ansible, SaltStack, NAPALM) I enabled 'ipfw' mode in Suricata which allows it to act in Intrusion Prevention Mode with divert(4).
Last change of note is finally having finished a diff to provide a single interface in the kernel to provide entropy to the randomness subsystem. Previously there were different functions depending on the source, i.e. add_timer_randomness(), add_mouse_randomness(), etc. It is no longer useful to distinguish where the data comes from, on account of having so many good sources all throughout the system, it was only adding complexity these days. So they were all replaced with a single enqueue_randomess() and that's it.
All good things must come to an end, and so did this week fuelled by cider and crêpes. Thanks a lot to gilles@ for hosting another hackathon in Nantes (this time riot-and-strike-free) and to the OpenBSD Foundation and Epitech for funding and hosting.
Thank you, Jasper, for your report and your work in Nantes!
(Comments are closed)