This can be used to make decisions about whether or not we should assign a virtual status to directories, prune them from a status display, or similar.
Subversion itself is roughly 30% faster if we do not ask for status of all files, the XML processing though is magnitudes faster, since it will only have to deal with a few records (in the general case) rather than a record per file (xsltproc seems to do around ~10,000 files per second on my Mac Pro, though I have no idea if it scales linearly).
The scm::status::versioned status has now been removed.
The upshot of this is that SCM drivers can leave out files that are “clean” from their status map, and TextMate will assume they have status “none” (clean). This cuts down on the amount of data that needs to be sent (and/or processed) by the SCM drivers.
For a repository with 7,767 files it takes 3.5 seconds for Python to convert a full file listing from XML to the simpler line-based format. Using xsltproc the same takes 0.74 seconds.
According to issue #340 TextMate “freeze for up to about 5 - 10 minutes” when doing svn status. Not sure how much this change helps on that, no doubt it’s an improvement.
This is done by overloading performKeyEquivalent: for the main menu (by setting it to OakMainMenu in MainMenu.xib).
The default implementation of performKeyEquivalent: does not know about scope selectors so it will sometimes call the wrong actions. For example a bundle overloading ⌘X for “-dyn.selection” may still see the item fire with a selection.
This wasn’t a problem previously, as we set the key equivalents via (deprecated) Carbon API (so presumably Cocoa didn’t know about the keys).
This is a preliminary implementation, I just want it out there to have it tested for edge-cases. Long-term we can probably remove the overload of performKeyEquivalent: on OakTextView, which sort of does similar (stops Cocoa from executing the wrong menu actions or interfering with multi-stroke key bindings).
This allows the SCM drivers to not report status for non-modified versioned files.
The problem was previously that if a driver didn’t report status for such files, having a file go from modified → clean wouldn’t always cause a refresh, since no change notification was generated for the path.
For the svn driver there is a 50% overhead by svn itself in reporting status for files without status, this overhead is likely magnified by the following XML step (where XML for all files in the working copy needs to be parsed).
The recent move to 64 bit broke the (binary) protocol used between TextMate and the server for authenticated reads/writes (since size_t is now 64 bit).
This fixes issue #354.
Since we now use the (numeric) user ID as part of TextMate’s named socket (to allow multiple users on the same system to each run an instance of TextMate), mate needs to know this.
Running ‘mate’ via ‘sudo’ has getuid() return zero (root), hence why we need to specifically handle this.
The text encoding (or even field type) can be practically anything, so we need to either handle everything or coerce to UTF-8.
This is a quick fix for an issue with QuickCursor, a more thorough review of encoding used with AppleEvents is to follow.
Turns out this is required in the situation where the document occupy less space than the visible document rect.
In this case, inserting new lines will not resize the text view, since it has been artificially enlarged to fit the visible document rect, therefore the gutter view will not update properly when inserting new lines.
I think there is a better solution than have the text view refresh the gutter, reverting is just meant as a quick fix.
This reverts commit 4e2ae3c508.
This is a fairly subtle bug: We were implicitly creating a temporary std::string object and requesting c_str from this object.
This C string’s lifetime was that of the temporary std::string, and as we did a lot of std::string creation in the following line, with the new libc++, there was a good chance that the temporary string buffer would be re-used and so, our working directory would be set to a bogus value.
This closes issue #329.
Renamed the DialogPopupMenuTarget class to avoid a run-time warning logged in Console.
Users will sometimes quote stuff from Console in bug reports, so let’s cut down the noise.
1. Default SDK set to 10.8 (since we now use subscript operators).
3. Version is tagged (by default) with “-git”. Not too happy about this, but I need a way to filter out “build from source” when analyzing crashes and similar. I’ll revise the solution later.
2. Revision only taken from api.textmate.org (no git fallback).
4. Only multimarkdown is a valid markdown compiler (due to recent changes in `bin/gen_html`).
I suspect that this was added because some operations were not properly detected, though the only thing I can find, that doesn’t properly update the gutter, is when deleting folded code in a way that doesn’t change the bounds of the text view. This code however does not address that issue.
Incase the last line of a file is folded (has no newline) then the gutter would first receive a row record for the line starting this fold, and then when querying for the next row, it would get the last line of the layout, this had a different line number but shared same coordinates as the previously rendered line, so the gutter would draw on top of that.
Closes issue #256.
There is a somewhat frequent crash caused by infinite “did resize” notifications. It’s unclear exactly why this happens, but basically a subview shouldn’t resize itself unless asked, and it shouldn’t resize siblings either.
We now have an explicit sizeToFit which takes care of sizing the gutter view. It is the responsibility of the owner (of this gutter view) to resize the gutter’s parent scroll view and any siblings that may need adjustment (after calling sizeToFit).
Additionally it is the responsibility of the owner to call sizeToFit when there is a chance the gutter view size needs updating. Generally though, the owner should know, as it is the data source for everything.
Previously we would lock focus on our source image. This is bad because we do rasterization (to a fixed pixel format / density) and it (potentially) destroys / mutates the source image.
We now use a second (new) image for the compositing. This is still not ideal, as this second image still is “off screen” so lack the proper graphics context during rendering. A better approach might be to create a CGImage using the graphics context and use that as an (image) mask and do a stroke, but this would require the images to be changed and I am not really certain it will work.
I am seeing crash reports from resolve_links and I can’t think of any other reason for this crash, than readlink() returning a value which is neither -1 nor in the range [0..len].
We now leave a potential trailing slash in the result.
This fixes the issue of the missing slash in ⌘T — while I normally prefer the slash to be stripped, I think path::with_tilde should be an exception since effectively it is just replacing the path prefix, it shouldn’t do any real processing (although we do actually normalize the string).
Also fix a bug where we would use tilde for paths pointing to a user folder for a user with current user’s name as prefix.
