On Sun, Jul 21, 2024 at 08:18:50AM +0200, Paul Gevers wrote: >... > PS: one other idea I'm having. There are multiple teams doing these kind of > rebuilds and archive creation; does each have their own tools and does it > their own way, I guess so? Has anybody ever tried to have the teams join > forces? ruby, perl, python and ocaml I already know of. (I fear I'm hearing > PPA/bicksheds/debusine resonating in my question). >...
The proper solution would be that release team tooling automatically schedules rebuilds and autopkgtests with the rebuilt packages for transitions prepared in experimental. ocaml is special since it is not an auto-* transition, but for auto-* transitions it feels pretty archaic that manual work is required for anything - every (re)upload of an auto-* transition package should start a CI run that does everything you are currently asking people to do manually. cu Adrian BTW: Don't assume that tooling would be the only issue, and that all DDs would have hardware capable of doing rebuilds. Some DDs might have more RAM than others have diskspace. Many DDs appear to have desktops/laptops that are badly equipped for rebuilds involving larger packages. "Out of memory; g++ killed (a machine with > 4 GB needed)" from the gsl transition does not sound like a machine where anyone should even attempt rebuilds of larger packages. Trying to build larger C++ sources with less than 4 GB RAM per virtual core can be a horrible experience. The maximum amount of RAM that fits into < 200 Euro motherboards today is 256 GB. You might not know whether the person you are asking to do a rebuild has a desktop with 256 GB RAM, or a laptop with a 128 GB SSD.
