The motivation for this change is that we may enable MallocDebug when running tests, which is done via environment variables that we do not want to be passed on to the shell tools we execute.
No file descriptors (other than input/output) should be inherited by our child process. We set FD_CLOEXEC ourself, but system frameworks might not always set it, or there might be a race condition (e.g. with socket(), pipe(), mkstemp(), etc.).
Failure to create a pipe does happen “in the wild” (with the error “too many open files”). Since majority of subprocess executions relate to gathering version control info we can probably ignore the failure, as not having version control info is better than crashing.
OTOH that pipe() fails does seem to be a symptom of a more serious issue.
The io::spawn implementation has also been updated to use io::create_pipe which sets FD_CLOEXEC, so there should be no reason to set this.
Since Mercurial is not using a unique “path suffix” we need to handle this via the default properties (instead of in the grammar’s file type extension array).
The latter is updated via a dispatch queue and thus might be delayed (to after end of current event loop cycle). This is relevant when WebView itself terminates the URL loading and calls the “WebView should close” delegate method.
Here the HTML output view property has been updated (via the WebView delegate method) but the command runner has not yet executed the completion code (which updates it’s “is running” state).
This can be set to ‘onDocumentChange’ to indicate that the command should be re-run when the document changes, though the code hasn’t yet been written to handle this.
We want to disable “make selection visible” when the user is drag-selecting, though the previous check was too broad and would disable it if user opened a document via double click (e.g. a find in folder result).
Previously we only updated the system pasteboard if it was different, but some users copy rich text from one application, paste it in TextMate and then copy it (in TextMate) with the expectation that the rich styles are stripped.
This ensures that ⌘Z / ⇧⌘Z are only targeting the file browser when it has focus.
Previously these keys would act on the file browser undo stack when a view without it’s own undo/redo stack had focus, e.g. HTML output view.
Since it is the cell that is providing the accessibility attributes the simplest way I could think of to have it ignored was to remove NSImageView’s support for the NSAccessibilityChildrenAttribute.
I tried using accessibilitySetOverrideValue:forAttribute: but it did not work for me, so a subclass was required.
While code reuse is an added benefit, the main motivation was to fix the appearance of the action popup button in the Find window. The first (placeholder) item in the Find window's action button was assigned a nop action so that when the button was clicked and it's menu validated, the "action" image would always appear as disabled.
Previously it would just make sure the selection was visible, so if the symbol was already visible, no scrolling/centering would take place.
For most actions we want to minimize scrolling since it makes it easier to keep track of the location in the document, but for the “Go to …” actions, we’re doing an absolute positioning rather than a relative jump, so it makes sense to always center.
For bookmarks, it’s only when choosing a specific bookmark from the Navigation menu that we center selection, not when using the next/previous bookmark actions.
The state being restored here would be file browser state, which might confuse the user if they are opening a folder via “mate” and then get a file browser restored for that project folder that show another location than the actual folder they open.
The reason we ignore the state when loading, rather than not saving it in the first place, is that by saving the state we make the folder show up in File → Open Favorites…
Previously we would set a mark, mutate the buffer, and then restore selection based on how the mark was moved.
When a mutation was changing the buffer to its existing value (e.g. toggling case of a space character) then the buffer would effectively ignore the mutation, and the mark would not be properly updated, meaning for “toggle case of character” the caret would not be moved forward.
Additionally we previously always moved to the end of a mutation, even if it was done on a user selection. We now only move to the end if the user didn’t select anything (but ran the mutation on an implicitly selected unit, like a word). The “toggle case of word” action is affected by this.
