Danny Milosavljevic <dan...@friendly-machines.com> writes: >> I think we should revert that commit until we can revisit. It has >> caused enough long-standing issues and the gain of having it appears small. > > I agree. > > Though if the store key of grafts is good enough so that we could just > fine-grain ALL grafts that would work too. > > I think the reason is that for dependencies, we have 1-output grafting going > on each, but for other things we have many-output grafting going on each. > > This has caused enormous problems: Approximately nobody programs GNOME > programs in C anymore. And the official GNOME bindings are all gobject > introspection. But that means as long as any of this grafting thing is going > the way it is now, no non-C GNOME programs work correctly (since as soon as > you use Gtk and glib, which are different packages (which will mess up > gobject introspection typelibs) you have this problem). > > It could conceivably be fixed in either direction: > > 1. Make all grafting 1-output, XOR: > 2. Make all grafting all-output. > > (I don't understand the key determination of grafts well enough to know > whether 1 would work) > > I mean it has made me finally spend a lot of time learning emacs because > literally nothing else that has a Gtk GUI would work right. I don't think > that's a good way ;) > > In any case, until we do know what to do, let's revert to a known-good state, > which is 2. > > I would really like 1 to work, though. > > Which to choose probably has a computation (lots in 1) vs storage (lots in 2) > trade-off.
Hi Danny, I have good reason to suspect this same issue is causing my GNOME services to crash, and those are pretty important! For example, my laptop will not lock when I close it anymore, because the Screensaver service cannot start! If we can revert to a known-good state, let’s do it as soon as possible please! I think this should be considered a very important bug. Have a nice day, Noé
signature.asc
Description: PGP signature
