Contributed by Peter N. M. Hansteen on from the cooking up the packages dept.
espie@), who writes:
I already came to Bucharest a year ago for EuroBSDcon, but I welcomed the chance at spending more time here, especially at a hackathon organized by Paul, who is such a great guy.
I heard that there was a lot of chanting involved around the city, but we had magical weather, totally unseasonally warm and sunny for november in Romania.
After an uneventful trip from Paris, I located the hackroom, inside Bucharest's central university, definitely a building with personality... Reminded me quite a bit of the Sorbonne, between the classy building you're not allowed to repair because it's an historical monument, the labyrinthine corridors, and the crazy night guards.
I was very happy to rejoin the crew, a lot of friends I made over the years, along with some new friends
rsadowski@, whom I hadn't met in person yet. Sadly, neither
jasper@could make it this year. We miss you guys.
The best laid plans usually fall apart during the first days of a hackathon, but surprisingly, I managed to keep tweaking dpb for a few days, documenting quickly written code, changing stuff that made no sense (and often discovering that it actually made sense) and fixing some actual bugs (the error path for awol hosts was actually quite amiss). I think I only broke the tree twice (sorry
naddy@)! The large plan was (and still is) to integrate portroach capabilities inside dpb proper, since the tool already scans the full ports tree, and has extensive job running/monitoring capabilities.
During the long testing times, I tinkered a bit with some future ports, and derailed Ingo (
schwarze@)'s plans... pretty nasty roff in a subsidiary port that uncovered several misconceptions of mandoc... I guess there's a first time for everything, but this got me my first "thank you for the bug" ever.
Then the plans got derailed. People started talking about having debug packages, and
pirofti@actually had a plan. Turns out the gdb from ports can team up with our objcopy to move debug information on the site, so that it can be packaged separately (somewhat nasty .debug/ directories everywhere, but at least it worked out of the box). We (=more or less
aja@... and some others) decided to make this "opt-in", because it's likely the size increase of snapshots might break the mirrors eventually, so let's start small.
The first draft had a lot of manual interaction, but it got ready really fast, without too much infrastructure churn (just hide the debug package creation along the normal package creation, give the list of files to process by hand, and nothing else matters), and so we could check that it works.
aja@was really quick to test it on some gnome stuff, and
landry@managed to get firefox debug packages up surprisingly fast, considering the size of the monster.
The devil lives in the details, so I spent a few days polishing that work, and I think we now have stuff in fairly good shape. It was definitely serendipitous, because I couldn't have done things that way a year ago, as I did most of the heavy lifting during the rewrite of update-plist: this allowed me to clone the tool into build-debug-info, which *also* reads all existing packing-lists from all subpackages, and creates the corresponding debug packing-lists and packing-info generation dynamically.
I tried to apply agile development methodologies with great success. Specifically, people were yammering for fully automated infrastructure from the start, but I took the time to look through the process before I decided it could work automatically.
Along the way, packing-lists gained some new annotations: we already annotate binaries and shared libraries, but full debug requires annotating other shared objects as well, and static libraries (we cope with static libraries by providing the full library with debug info in the debug package instead of the normal package).
Paul was a wonderful host! and Bucharest has a lot of great restaurants/bars, so we really enjoyed the romanian cuisine, and also some more international food. As usual, we had lots of future plans and stupid discussion over drinks. The OpenBSD project would be totally different without the bonding that happens at those times.
On monday, we went to the traditional hackathon "BBQ". Paul had insisted that we spend the day at Thermes, a recent spa/pool/aquatic amusement park located in the countryside close to Bucharest. I believe that everyone who went had a blast. Felt like a quick trip to the Caribean, what with the hot baths and wonderful weather, and we had a lot of fun with the rides. Two thumbs up.
So, now I'm writing this report on the plane back to France. The arrival is going to be tough, as France is enjoying its usual shitty november weather and I'm pretty sure the frowning locals are going to make me regret leaving Romania very quickly.
Thanks again to Paul and the fundation for organizing this.
Thanks for the report and all the good work, Marc!
(Comments are closed)