In my app, a graph in one view can be dragged to a second view so that the new graph replaces the second view (like a copy/paste effect with a drag and drop feature). The app works if the delegate protocol is taken out so that the second view handles the change in function itself. When the protocol is added, the app crashes in the main file at
return UIApplicationMain(argc, argv, nil, NSStringFromClass([Load_CreatorAppDelegate class]));.
There isn't any error output other than the standard (lldb). Even when I take out the call to the delegate (keeping in the code), the app crashes. I know that it has to be related to the protocol code, though, because it worked fine before that.
Here is part of the code for the second view (BeamView):
[self drawSupportsAtLeftPoint:self.beamBottomLeft rightPoint:self.beamBottomRight inContext:context :leftPin :rightPin];
BOOL pt = NO;
if (self.tempLoad) {
//self.loadGraph = [self.dataSource changeToTempLoad:self]; NOTE #1
//if (self.tempPtLoad.x != 0 || self.tempPtLoad.y != 0) pt = YES;
pt = [self changeLoad];
[self drawLoadWithFunction:self.loadGraph inContext:context fromPoint:self.beamTopLeft toPoint:self.beamTopRight withAlpha:0.3 isPointLoad:pt inBlack:YES];
}
else {
self.loadGraph = ^(int x) {return x/15;};
[self drawLoadWithFunction:self.loadGraph inContext:context fromPoint:self.beamTopLeft toPoint:self.beamTopRight withAlpha:1 isPointLoad:pt inBlack:NO];
}
self.tempLoad = NO;
NOTE #1: These lines that are commented out are the ones that call on the delegate. Those two methods and their implementation are the only changes I made.
I'm completely confused, any help would be greatly appreciated. What are possible reasons the app will crash in the main file?
Ok! I feel kind of stupid, but it turned out that the crash didn't have anything to do with delegation (well, kind of). I had deleted outlets in the ViewController.m file without disconnecting them in IB, which was causing the crash.
I'd forgotten that I'd done that, and so it took me a while to think of it--it wasn't until I went back to an older saved version that I saw the differences.
Related
"[GCController controllers]" does not contain any controllers that were connected prior to application launch
TLDR;
I am trying to implement gamepad input on macOS using the Game Controller Framework. When invoked in my code, [GameController controllers] always returns an empty list until new controllers are connected. It never reflects gamepads connected to macOS prior to application launch, except if you disconnect them and reconnect them while the app is running. Does anyone know what I need to do to make controllers populate with pre-launch connections?
Full question
Now that Apple has added support for Xbox and Playstation controllers to the GameController framework, I'm trying to use it for gamepad input on a C++ game engine I'm developing. I'm using the framework instead of IOKit in order to "future-proof" my games to support additional controller types in the future, as well as to simplify my own input handling code.
Like many other game engines, I've foregone using NSApplicationMain() and nib files in favor of implementing my own event loop and setting up my game window programmatically. While my "Windows style" event loop appears to be working correctly, I've discovered that [GCController controllers] does not. The array it returns is always empty at launch, and will only ever reflect controllers that are connected while the game is running. Disconnecting a pre-connected controller does not trigger my GCControllerDidDisconnectNotification callback.
Here is a simplified version of my event loop:
int main(int argc, const char * argv[])
{
#autoreleasepool
{
// Create application
[NSApplication sharedApplication];
// Set up custom app delegate
CustomAppDelegate * delegate = [[CustomAppDelegate alloc] init];
[NSApp setDelegate:delegate];
// Activate and launch app
[NSApp setActivationPolicy:NSApplicationActivationPolicyRegular];
[NSApp setPresentationOptions:NSApplicationPresentationDefault];
[NSApp activateIgnoringOtherApps:YES]; // Strictly speaking, not necessary
[NSApp finishLaunching]; // NSMenu is set up at this point in applicationWillFinishLaunching:.
// Initialize game engine (window is created here)
GenericEngineCode_Init(); // <-- Where I want to call [GCController controllers]
NSEvent *e;
do
{
do
{
// Pump messages
e = [NSApp nextEventMatchingMask: NSEventMaskAny
untilDate: nil
inMode: NSDefaultRunLoopMode
dequeue: YES];
if (e)
{
[NSApp sendEvent: e];
[NSApp updateWindows];
}
} while (e);
} while (GenericEngineCode_Run()); // Steps the engine, returns false when quitting.
GenericEngineCode_Cleanup();
}
return 0;
}
I've confirmed that even when using [NSApp run] instead of [NSApp finishLaunching], the behavior is the same. As best as I can tell, the problem is that there's something NSApplicationMain() does that I'm not doing, but that function is a black box -- I can't identify what I need to do to get controllers to populate correctly. Does anyone know what I'm missing?
The closest thing I could find to an explanation of this problem is this answer, which suggests that my app isn't getting didBecomeActive notifications, or that at the least, the private _GCControllerManager isn't getting a CBApplicationDidBecomeActive message. I'm not a professional macOS developer, though: I don't know if this actually applies to my situation, or how I'd go about correcting the problem if it does.
After a huge amount of time searching, I found the answer on my own. It turns out that my code wasn't the problem -- the problem was that my Info.plist file was having its CFBundleIdentifier value stripped out due to a problem with my build system. It appears that the Game Controller Framework needs the bundle identifier to correctly populate [GCController controllers] at launch. While a missing CFBundleIdentifier would have been a problem anyway, as a Windows person it didn't occur to me that the identifier might be used for things besides the App Store, so I let it slide until now.
If someone else has this problem, make sure that CFBundleIdentifier isn't missing or empty in Info.plist in your assembled app bundle. In my case with Premake, I had to manually set PRODUCT_BUNDLE_IDENTIFIER with xcodebuildsettings so that $(PRODUCT_BUNDLE_IDENTIFIER) would get properly replaced in Info.plist.
I have been tinkering with a UICollectionViewLayout for a few days now, and seem to be stuck on getting a Supplementary view to delete with animation using the finalLayoutAttributesForDisappearingSupplementaryElementOfKind:atIndexPath:
I have finally made some headway and discovered this issue. Initially it looked as if it was not animating, because i was not getting a fading animation. so i added a transform on it to get it to rotate... and discovered a supplementary view is in fact being animated, but another one just sits there. until its finished and the unceremoniously disappears not being affected by being told to disappear at all. So to sum up it looks like an exact copy is made and that is animated out for finalLayoutAttributesForDisappearingSupplementaryElementOfKind:atIndexPath: and the original just sits there unmoving and animating, then both disappear.
This copy or whatever it is does not disappear either until i scroll the view.
Does anyone have any insight into this issue?
Update
So after some tweaking, experimentation and many hypothesis later as to what was causing the issue, it turns out that the culprit was this method
- (NSArray *)indexPathsToDeleteForSupplementaryViewOfKind:(NSString *)kind
{
NSMutableArray *deletedIndexPaths = [NSMutableArray new];
[self.updatedItems enumerateObjectsUsingBlock:^(UICollectionViewUpdateItem *updateItem, NSUInteger idx, BOOL * _Nonnull stop)
{
switch (updateItem.updateAction)
{
case UICollectionUpdateActionInsert:
{
if (_unReadHeaderIndexPath != nil && [self.layoutInformation[ChatroomLayoutElementKindUnreadHeader] count] == 0)
{
[deletedIndexPaths addObject:_unReadHeaderIndexPath];
_unReadHeaderIndexPath = nil;
}
}
break;
case UICollectionUpdateActionDelete:
{
[deletedIndexPaths addObject:updateItem.indexPathBeforeUpdate];
}
break;
default:
break;
}
}];
return deletedIndexPaths;
}
The part to focus on is here
if (_unReadHeaderIndexPath != nil && [self.layoutInformation[ChatroomLayoutElementKindUnreadHeader] count] == 0)
{
[deletedIndexPaths addObject:_unReadHeaderIndexPath];
_unReadHeaderIndexPath = nil;
}
Particularly the _unReadHeaderIndexPath = nil;
What this is doing is making a supplementary view disappear which indicates unread messages when the user inserts any text and then setting it to nil so it is not added multiple times, the logic is sound however the method in which is employed seems to be the issue.
indexPathsToDeleteForSupplementaryViewOfKind: is called multiple times and one can assume it needs multiple passes through it to work properly but since we nil _unReadHeaderIndexPath; it seems like it isn't added to the returned array at the appropriate time.
It also seems like the incorrect method in as I'm checking for insertion action in a method name indexPathsToDeleteForSupplementaryViewOfKind: so the solution was to move all of that logic to the method
prepareForCollectionViewUpdates:
Where we originally set the values for self.updatedItems in the first place. Which means i can get rid of that member variable altogether. And it animates and disappears without any odd copy sticking around.
I was under the impression that when using bindings (been following this tutorial despite being outdated. http://cocoadevcentral.com/articles/000085.php - You can use it to see what I'm doing) the Persistent Store would automagically save the changes you make. In fact, though it was hours ago and I wouldn't be surprised if I'm now going mad, I got it working, and when I made a change it would persist on rebuilding the app.
However, the test app I've built following the tutorial no longer saves and despite showing the changes I make within the app, they disappear once I re-run the app. I've been checking the Core Data debug menu and nothing happens when I press the "+" button which is set up to the "Add" method of my NSArrayController. I know it's accessing my data model too as my textField for the Title (again, see the tutorial so you know what I'm referring to) adopts the default text I put in the DataModel section. The only thing missing therefore is the actual saving.
So my real question is, based on the tutorial, what part of the bindings actually makes the managedObjectContext save? Is there a flag or something that isn't checked?
I don't know if it's important or not, but there were differences between the tutorial and my project, mainly that the NSArrayControllers are bound to "App Delegate"with a Model Key Path of "self.managedObjectContext". Also, I removed all the relationships in an attempt to whittle down the issue.
Any help would be greatly appreciated.
Regards,
Mike
UPDATE: Here are some pictures that show the bindings.
How I set up the NSArrayController:
Here is how is how my Data Model Looks:
Lastly, this is how I set up the TextFields to update the NSArrayControllers:
I hope this helps to get a an ideas as to the set up.
Thanks,
Mike
Could you check to make sure you've copied all the Core Data boiler-plate code from the source code of the tutorial you mentioned.
Specifically this part in the App Delegate:
- (NSApplicationTerminateReply)applicationShouldTerminate:(NSApplication *)sender {
NSError *error;
NSManagedObjectContext *context;
int reply = NSTerminateNow;
context = [self managedObjectContext];
if (context != nil) {
if ([context commitEditing]) {
if (![context save:&error]) {
// This default error handling implementation should be changed to make sure the error presented includes application specific error recovery. For now, simply display 2 panels.
BOOL errorResult = [[NSApplication sharedApplication] presentError:error];
if (errorResult == YES) { // Then the error was handled
reply = NSTerminateCancel;
} else {
// Error handling wasn't implemented. Fall back to displaying a "quit anyway" panel.
int alertReturn = NSRunAlertPanel(nil, #"Could not save changes while quitting. Quit anyway?" , #"Quit anyway", #"Cancel", nil);
if (alertReturn == NSAlertAlternateReturn) {
reply = NSTerminateCancel;
}
}
}
} else {
reply = NSTerminateCancel;
}
}
return reply;
}
If it's there, changes will be saved when the app is terminated normally. Pressing the 'stop' button in Xcode will terminate the app immediately, without going through the method mentioned above.
My guess is that you are not going mad, but first exited the app properly and have been pressing the 'stop' button later ;).
I have an NSOpenPanel with an accessoryView; in this view the user chooses a couple of radio button to change the allowed types. When the panel opens, the right files are enabled, the other disabled. Ok, good.
Now the user changes the radio buttons, the viewController of the accessoryView observe the changes in the radio button matrix and changes consequently the allowedTypes of the NSOpenPanel.
After that, following Apple documentation, it calls -validateVisibleColumns, but nothing visible changes in the panel. That is: the right files seems disabled: I can choose them but they are in grey!
Another wrong effect: I select a file (enabled), change the file type, the (now wrong) file remains selected, with the OK button enabled: but this is the wrong file type! It seems that the change happens but the interface doesn't know!
My code is (selected is bound to the matrix of radio button):
- (void)observeValueForKeyPath.....
{
NSString *extension = (self.selected==0) ? #"txt" : #"xml";
[thePanel setAllowedFileTypes:#[extension, [extension uppercaseString]]];
[thePanel validateVisibleColumns];
}
I first tried to insert a call
[thePanel displayIfNeeded]
then I tried with
[thePanel contentView] setNeedsDisplay]
with no results. I also tried to implement the panel delegate method panel:shouldEnableURL:, that should be called by validateVisibleColumns: I just found that it was called just once, at the opening of NSOpenPanel.
Can someone have an idea why this happens? I tried all this with sandboxed and not-sandboxed applications, no difference. I'm developing on ML with 10.8 sdk.
Edit
By now the only way to avoid the problem is to implement panel:validateURL:error, but this is called after the user clicked 'open' and it's very bad.
I have the exact same problem, under 10.9, non-sandboxed, and have spent the better part of this DAY trying to find a solution!
After A LOT of tinkering and drilling down through the various classes that make up the NSOpenPanel (well NSSavePanel really) I did find a way to force the underlying table to refresh itself:
id table = [[[[[[[[[[[[_openPanel contentView] subviews][4] subviews][0] subviews][0] subviews][0] subviews][7] subviews][0] subviews][1] subviews][0] subviews][0] subviews][0] subviews][2];
[table reloadData];
Of course, the best way to code this hack would be to walk down the subview list ensuring the right classes are found and eventually caching the end table view for the subsequent reloadData calls.
I know, I know, this is a very ugly kludge, however, I can not seem to find any other answer to fix the issue, other than "file a bug report". Which, from what I can see online people have been doing since 1.8! :(
EDIT:
Here is the code I am now using to make my NSOpenPanel behave correctly under 10.9:
- (id) openPanelFindTable: (NSArray*)subviews;
{
id table = nil;
for (id view in subviews) {
if ([[view className] isEqualToString: #"FI_TListView"]) {
table = view;
break;
} else {
table = [self openPanelFindTable: [view subviews]];
if (table != nil) break;
}
}
return table;
}
- (void) refreshOpenPanel
{
if (_openPanelTableHack == nil)
_openPanelTableHack = [self openPanelFindTable: [[_openPanel contentView] subviews]];
[_openPanelTableHack reloadData];
[_openPanel validateVisibleColumns];
}
This code requires two instance variables _openPanel and _openPanelTableHack to be declared in order to work. I declared _openPanel as NSOpenPanel* and _openPanelTableHack is declared as id.
Now, instead of calling [_openPanel validateVisibleColumns] I call [self refreshOpenPanel] to force the panel to update the filenames as expected. I tried caching the table view when the NSOpenPanel was created, however, it seems that once you "run" the panel the table view changes, so I have to cache it on the first update instead.
Again, this is a GIANT hack, however, I do not know how long it will take Apple to fix the issue with accessory views and the file panels, so for now, this works.
If anyone has any other solutions that are not huge kludges please share! ;)
An implementation in swift of Eidola solution.
Biggest difference is that I search for a NSBrowser (sub)class rather than a specific class name. Tested on 10.10 (not sandboxed).
private weak var panelBrowser : NSBrowser? //avoid strong reference cycle
func reloadBrowser()
{
if let assumedBrowser = panelBrowser
{
assumedBrowser.reloadColumn(assumedBrowser.lastColumn)
}
else if let searchResult = self.locateBrowser(self.panel?.contentView as! NSView)
{
searchResult.reloadColumn(searchResult.lastColumn)
self.panelBrowser = searchResult //hang on to result
}
else
{
assertionFailure("browser not found")
}
}
//recursive search function
private func locateBrowser(view: NSView) -> NSBrowser?
{
for subview in view.subviews as! [NSView]
{
if subview is NSBrowser
{
return subview as? NSBrowser
}
else if let result = locateBrowser(subview)
{
return result
}
}
return nil
}
Edit:
Ok, so the code above will not work all the time. If it's not working and a file is selected (you can see the details/preview), then you have to reload the last to one column instead of the last column. Either reload the last two columns (make sure there are at least 2 columns) or reload all columns.
Second problem: if you reload the column, then you lose the selection. Yes, the selected files/directory will still be highlighted, but the panel will not return the correct URL's.
Now I am using this function:
func reloadBrowser()
{
//obtain browser
if self.panelBrowser == nil
{
self.panelBrowser = self.locateBrowser(self.panel?.contentView as! NSView)
}
assert(panelBrowser != nil, "browser not found")
//reload browser
let panelSelectionPatch = panelBrowser.selectionIndexPaths //otherwise the panel return the wrong urls
if panelBrowser.lastColumn > 0
{
panelBrowser.reloadColumn(panelBrowser.lastColumn-1)
}
panelBrowser.reloadColumn(panelBrowser.lastColumn)
panelBrowser.selectionIndexPaths = panelSelectionPatch
}
Just upgrade to xcode 6.3 (on Yosemite 10.10.3) and its ok. Apple fixed the bug (no more need Eidola Hack).
Important update: I found out that most part of my question was based on a false premise (see my answer below). Notifications actually got to the receiver, they just got there too fast. (Although, it still doesn't explain why the behavior with breakpoint and without it was different.)
I'm developing the app that calculates the hashes of files given to it. The calculation takes place in SHHashComputer. It's an abstract class (well, intended to be abstract, as there are no abstract classes in Objective C) that takes the file path and creates an NSInvocationOperation. It, in turn, calls the method (void)computeAndSendHash, which uses the file path saved in the object to compute hash and sends it as notification. The actual computing takes place in (NSString*)computeHash method that child classes need to override.
Here's SHHashComputer.m:
- (NSString*)computeHash {
return [NSString stringWithFormat:#"unimplemented hash for file %#", self.path];
}
- (void)computeAndSendHash {
NSString *result = [self computeHash];
NSString *notificationName = [NSString stringWithFormat:#"%#%#",
gotResultNotification,
self.hashType];
[[NSNotificationCenter defaultCenter] postNotificationName:notificationName
object:result];
self.operation = nil;
}
And here's SHMD5Computer.m (the child class of SHHashComputer):
- (NSString*)computeHash {
return #"MD5 test"; // it actually doesn't matter what it returns
}
I won't bother you with the receivers of notification. Let's just say that as long as I comment out the computeHash method in SHMD5Computer.m everything works just fine: the notification with text "unimplemented ..." is received & displayed in GUI. But if I don't — then it gets really interesting.
If I don't set up any breakpoints, the notification just never comes. However, if I set up a breakpoint at the declaration of computeHash in SHMD5Computer.h and then step over until the line 'self.operation = nil', and continue execution at that point, the notification gets to destination. If I don't stop there, the debugger suddenly switches to the state as if it isn't debugging anything, and the app freezes.
I don't think that 'WTF' is a good form for a question here, so let me put it this way: am I missing something? Are there errors in my code? What can cause this type of behavior in xcode? How can I fix this?
(If you'll want to get all my code to reproduce it, I'll gladly give it to you.)
More experiments:
If I continute execution exactly after stopping at breakpoint, the application encounters EXC_BAD_ACCESS error in the code that receives the notification, at the last line:
id newResult = [newResultNotification object];
if (newResult == nil)
[NSException raise:#"No object"
format:#"Expected object with notification!"];
else if (![newResult isKindOfClass:[NSString class]])
[NSException raise:#"Not NSString"
format:#"Expected NSString object!"];
else
self.result = (NSString*) newResult;
[self.textField setStringValue:self.result];
When I tried to reproduce the previous experiment, something even stranger happenned. In my debug setup, I have two hash computer objects: one SHMD5HashComputer (which we're talking about), and one SHHashComputer (which, of course, produces the "unimpemented" hash). In all previous experiments, as long as app didn't crash, the notification form SHHashComputer always successfully arrived. But in this case, both notifications didn't arrive, and the app didn't crash. (All the steps are exactly the same as in previous one).
As Josh Caswell pointer out in the comments, I wasn't using the notifications correctly. I should've sent the object itself as notification object, as described in documentation. I fixed that, and I'm getting exactly the same results. (Which means that I fixed it correctly, because sometimes the notifications work correctly, and also that it wasn't the problem).
More updates:
The notification that I'm sending should arrive at SHHashResultViewController. That's how I create it and register for notification:
- (id)initWithHashType:(NSString *)hashType {
self = [self initWithNibName:#"SHHashResultView" bundle:[NSBundle mainBundle]];
if (self) {
[self setHashType:hashType];
}
return self;
}
- (void)setHashType:(NSString *)hashType {
[self.label setStringValue:[NSString stringWithFormat:#"%#:", hashType]];
_hashType = hashType;
NSString *notificationName = [NSString stringWithFormat:#"%#%#",
gotResultNotification,
_hashType];
[[NSNotificationCenter defaultCenter] addObserver:self
selector:#selector(gotResult:)
name:notificationName
object:nil];
}
Actually, the question was based on a false premise. I thought that notification never came through because I never saw the information displayed in the GUI; however, my error was in the code of controllers (not published there) which made possible the situation in which the GUI first got results of hash calculation and only after that got information about a new input — which resulted in deleting all the text and activating progress animation.