On Saturday December 02 2017 17:36:03 David Faure wrote: > "Why" is always hard, but your questions are only about "what", so that's > easy, the code answers that ;)
No, I think it was really about the why of the what, but thanks for confirming a few things. > The goal of what? That categorized and off-by-default debug message? No, why the code would be doing this: > So this is reference counted. Nice, heh? Sure, nifty. I thought it was doing something like that, but why would you want to reference-count this kind of feature? It seems that a simple off/on kind of reference counting (none, one-or-more if you prefer) should be enough here. Or maybe I should ask: why is there no way to remove a watch on a file completely? I can think of a scenario where multiple things in an app would be watching a given file so you wouldn't want to release the watch until the last "thing" releases its watch, unless someone deletes the file. > > As a result I feel I need to wrap each call to KDW::addDir(foo) with a check > > of KDW::contains(foo). > > No you don't need to do that. Let me be a bit more specific. Of course I don't need to do this if all I want is to set a watch. But if watches are ref-counted that suggests at least that you have to remove an item as many times as you've added it to the watcher. (This is the bit where I didn't stare very hard at the code at all.) So if I want to be able to remove watches without having to keep track of how many times I set one, I do need those wrappers. In my case watches are set by a single function which loads directories and can thus be called lots of times but they're basically unset only when directories are deleted (which usually happens only once ;)). R.
