NSFileManager only provide a move function which fails with “file exists” if we make case changes on a case-insensitive file system, therefor we now explicitly test for that error and use <x-man-page://2/rename> if the source and destination refer to the same file, but the path differ.
Fixes#769.
This is both to speedup file browser updates (as there is a bit of lag waiting for fs-events) and to work with file systems that doesn’t support fs-events (issue #745).
The reason for this is that NSTextField will draw itself in the “faded” state (for non-key windows) unless some of its parent views are marked opaque.
The previous problem with not having the parent views opaque (see da3a91a7bb) seems to have been solved by setting the autohidesScrollers property to YES.
The window border drawing seems to be done by the root content view so it’s best to keep it around (as it may also do other things that we are unaware of).
This shouldn’t be necessary, as we have a ‘retainedSelf’ instance variable that isn’t cleared until ‘[self.window close]’ (the last line of the accept: method), nonetheless I am seeing (rare) crashes from this method, and it appears to be reading the ‘self.window’ property in the last line, so adding this extra safety and will monitor crash reports.
This fixes a crash bug when menu validation is invoked before the controller has set up a document. Hardening the other methods is just a precaution (as we generally shouldn’t make assumptions about the state of public properties.
This is instead of going through ‘self’ and was done to harmonize the code. Since two of the three variables are C++ types we cannot use dot syntax (self.property) when we wish to call a member function on the result, but using [self property] is ugly and would require us to then also use [self setProperty:value] (for consistency), hence why I went with direct access.
Since the selectedTabIndex property is closely related to documents and selectedDocument, I applied the same changes to this property.
Previously we kept the children which had been expanded, as these had an associated record for tracking file system changes and SCM info.
However, when expanding we weren’t actually re-using the item from the cached record, so not only was this wasteful, but it caused a problem because we effectively created a new item but used the old record with the old item, and while the two items would compare is equal when using isEqualTo:, they didn’t have the same memory address, so the outline view would not consider them equal, and would thus not react on reload requests when using the older item.
This is only to ensure everyone goes through the class method, which is easier to search for (constructing an FSItem from a URL is a little expensive so we wish to get rid of it where possible).
Since the created image is a “dumb” object that can’t provoke redrawing itself, we require that SCM info is already cached for the repository containing the file for which we are creating an image. If not, there will be no SCM badge, so most users of this API should setup their own SCM info and recreate the image when SCM status changes.
When requesting SCM info for a path, we now always return the cached info for the first parent of the path, but if the first parent of the path is not the direct parent, we do a more thorough search (which involve disk access) in a background thread.
The latter is required because we may have nested repositories. E.g. asking for SCM info for /path/to/project/submodule/file will initially return the root project’s cached SCM info (if the submodule info hasn’t yet been added to the cache) but will then later invoke the callback with the submodule’s SCM info.
The FFWindowController class is an implementation detail where the Find class is the public class, so depending on the latter should make the code safe against potential rewrites.
This is instead of setting an environment variable. Since there already is a user defaults setting for whether or not we want an untitled document created at startup, it seems nicer to have mate set that, instead of effectively introducing an alias for the setting.
This is to be used when the application is in the background. Rather than completely disable updates, it might perhaps be better simply to increase the delay used to throttle updates.
Also change SCM implementation so that it doesn’t use objects with automatic storage, as the objects may be accessed from threads, which may run after objects with automatic storage has been destroyed.
