Swift doesn't convert Objective-C NSError** to throws

I have some Objective-C legacy code, that declares method like

- (void)doSomethingWithArgument:(ArgType)argument error:(NSError **)error

As written here https://developer.apple.com/library/ios/documentation/Swift/Conceptual/BuildingCocoaApps/AdoptingCocoaDesignPatterns.html

Swift automatically translates Objective-C methods that produce errors into methods that throw an error according to Swift’s native error handling functionality.

But in my project described methods are called like this:

object.doSomething(argument: ArgType, error: NSErrorPointer)

Moreover, it throws runtime exception when I try to use them like:

let errorPtr = NSErrorPointer()
object.doSomething(argumentValue, error: errorPtr)

Do I need something more to convert Objective-C "NSError **" methods to Swift "trows" methods?

Solution

Only Objective-C methods are translated to throwing Swift methods, which do return a BOOL (not lower-cased bool), or a nullable-object. (Tested with Xcode 11.7, and Swift 5 language.)

The reason is that Cocoa methods always use a return value NO or nil to indicate the failure of a method, and not just set an error object. This is documented in Using and Creating Error Objects:

Important: Success or failure is indicated by the return value of the method. Although Cocoa methods that indirectly return error objects in the Cocoa error domain are guaranteed to return such objects if the method indicates failure by directly returning nil or NO, you should always check that the return value is nil or NO before attempting to do anything with the NSError object.

For example, the Objective-C interface

@interface OClass : NSObject

NS_ASSUME_NONNULL_BEGIN

-(void)doSomethingWithArgument1:(int) x error:(NSError **)error;
-(BOOL)doSomethingWithArgument2:(int) x error:(NSError **)error;
-(NSString *)doSomethingWithArgument3:(int) x error:(NSError **)error;
-(NSString * _Nullable)doSomethingWithArgument4:(int) x error:(NSError **)error;
-(BOOL)doSomething:(NSError **)error;

NS_ASSUME_NONNULL_END

@end

is mapped to Swift as

open class OClass : NSObject {

    open func doSomethingWithArgument1(x: Int32, error: NSErrorPointer)
    open func doSomethingWithArgument2(x: Int32) throws
    open func doSomethingWithArgument3(x: Int32, error: NSErrorPointer) -> String
    open func doSomethingWithArgument4(x: Int32) throws -> String
    open func doSomething() throws
}

If you can change the interface of your method then you should add a boolean return value to indicate success or failure.

Otherwise you would call it from Swift as

var error : NSError?
object.doSomethingWithArgument(argumentValue, error: &error)
if let theError = error {
    print(theError)
}

Remark: At

https://github.com/apple/swift/blob/master/test/Inputs/clang-importer-sdk/usr/include/errors.h

I found that Clang has an attribute which forces a function to throw an error in Swift:

-(void)doSomethingWithArgument5:(int) x error:(NSError **)error
  __attribute__((swift_error(nonnull_error)));

is mapped to Swift as

public func doSomethingWithArgument5(x: Int32) throws

and seems to work "as expected". However, I could not find any official documentation about this attribute, so it might not be a good idea to rely on it.