Some targets were including headers from frameworks not specified in their link dependencies. For a clean build this could cause an issue because the header was not available at the time of building the target.
The updated link dependencies are also based on what a target’s tests require. Ideally tests would have separate link dependencies, but as we don’t want to maintain this manually, this will have to wait until the build system automatically handles link dependencies.
Currently the commit command uses constants from the CommitWindow framework but should actually not be linked with it. However, the optimizer will strip dead code, so it should not result in much if any difference in the resulting binary and does solve a build dependency issue.
This is consistent with Next/Previous Bookmark shortcuts bound to F2/⇧F2.
The problem with ⌘> and ⌘< is that these keys are used to cycle through windows on some keyboard layouts.
Discussion at issue #1316.
The reason is that if there are no bundles, we will extract the default set to Application Support, and in doing so, we call tar which implicitly sets up the command environment, which is re-used for all successive commands.
The problem is that loading a plug-in may add to the environment (e.g. setting the DIALOG environment variable), so if plug-ins are loaded after the call to tar (as would be the case for on a clean system), the DIALOG variable will be missing from the environment.
Kudos to René Schwaiger for discovering this issue.
The semantic class must be of the form: `theme.«group»[.«whatever»]*`.
If a theme lacks a semantic class, or the semantic class does not have a `theme.` prefix then the theme gets into the “unspecified” group.
In theory we could analyze the themes for dark/bright background, but manually classifying themes seems more flexible, for example some themes have a transparent background, which we could put into their own group.
This sets '' (empty identifier) in Global.tmProperties that switches to system
specified spelling language (usually Automatic by Default which is OS X
default). It serves as a way back to the default when user picks any other
language than default from the menu for current buffer (effectively for
documents's directory).
This is only required for the WebView which many bundle commands use to show online documentation and similar.
It’s impractical to maintain a whitelist of allowed domains.
This is required for storing boolean values which the new NSAppTransportSecurity setting must use. It does not work using `YES` or `1` to represent `<true/>`.
The problem is that when we change user defaults a notification is posted, and this runs on the thread that did the change, so our own “user defaults did change” notification callbacks are executed on a background thread.
The use case for this feature is when an interactive rebase is in
progress and be able to, in one command, amend the current commit
and continue with rebase.
By default, if the --rebase-in-progress flag is passed when the
"commit" command is invoked the commit button will be in a
"commit and continue" state. This is indicated by the title of the
commit button now being "Commit & Continue". If the option (⌥) key
is being hold down the state will revert back to the standard "commit"
state and the title will change to "Continue".
If the butten was clicked when it was in the "commit and continue"
state the "commit" command will print "TM_SCM_COMMIT_CONTINUE=1"
(without the quotes) to stdout as the last line. If the button was
clicked when it was in the standard "commit" state the output may
contain "TM_SCM_COMMIT_CONTINUE=0", or the "TM_SCM_COMMIT_CONTINUE"
string will be absent.
This was done in an attempt to prevent the warning saying “This file is set to build for a version older than the project deployment target.”
Unfortunately saving all the xibs did not get rid of the warning, even though tehy are all set to have 10.7 as their build target.
The latest xib format does however remove about 10.000 lines of XML.
Since we are doing a literal (byte-by-byte) subset match we need to have the filter string use the same unicode normalization form as the items it is matched against.
For favorites and the file chooser, the items are file system names, which default to the decomposed form.
Ideally the filtering algorithm would know about diacritics, but in practice the current solution should work for most actual scenarios, and is much simpler (thus faster).
The static `std::map` instance (named just `map`, on line 79) had its initializer template params in the wrong order, which kept it from compiling. Swapping them as per this PR lets `indent` build and run with no issues.
