Rene Gollent <anevilyak@xxxxxxxxx> wrote: > On Fri, Jan 15, 2010 at 2:01 AM, Axel Dörfler <axeld@xxxxxxxxxxxxxxxx> > > wrote: > > anevilyak@xxxxxxxxx wrote: > >> More Tracker refactoring: > >> * TrashWatcher now keeps icons in sync on all volumes. > > Does that really make much sense? > The reason I did so was because, since the Trash dir is now visible > from the root of every volume, it would look odd if you added an item > to Trash and only the one on /boot showed up with the "trash full" > icon while all the others showed empty (regardless of which volume > the > item was moved to trash on). Besides the general attribute issue with multi-user, maybe it would make more sense to use a standard folder icon for the trash, as you cannot know the contents of the trash of the other users. > > What were your reasons not to follow the symlink approach, btw? > > I'm just asking; I don't really mind, as it doesn't really change > > anything from the status quo, anyway. > The main issue I had with the symlink approach is it's quite > cumbersome as far as showing the correct context menu, since it means > that the mouse handling logic would need to resolve the model's link > to determine it's in fact pointing at the trash, which, afaict is not > a reversible op on the model, so I wanted to avoid side effects there. > Also it seemed odd to me to have what looks like a symlink but doesn't > allow you to do any of the usual symlink ops, so I went with this > approach instead. Also since this way doesn't require any actual file > to exist in the Desktop dir. At least in Haiku you could actually delete the Trash directory from your Desktop, though :-) Not that I would miss that. > > I'm not sure if uppercase "Trash" is the name we should choose, as > > we > > only have lowercase directories otherwise. > With respect to the uppercase name, the problem is that BPose adopts > the on-disk name of its Model, and since the desktop instance of > Trash > is essentially now a pose with a model pointing at /boot/Trash, that's > the name it will take, whichever case it has. I could possibly > override that with a Model subclass though, will need to research > that > a bit, as we'll want to be able to override the name anyhow for > localization reasons. Which could be a genuine Model feature then, I think. > If that's workable, should I modify > find_directory to make the on-disk name /trash? Ideally I'd really > rather just not have the trash dir visible from the volume root but > I'm not sure as to the best way to go about that for the moment > without hurting dir population performance and/or the caveats of > resurrecting the previous hidden pose approach. Not sure what you are referring here exactly. What possible performance problems do you see because of what change? > > That reminds me: our FAT file system isn't called "dos" anymore... > Will update that too, what does ours name itself as? "fat". Bye, Axel.