I'm investigating whether I can get better performance from asl_log than NSLog on iPhone/iOS (probably...) but I'm stuck at a point where it doesn't seem that asl's log output will show up in the System Console (as viewable by a number of apps like System Console, iConsole, etc). I know that I'm setting it up right since I open with ASL_OPT_STDERR, and I see the log entries in XCode when the device is tethered.
I've explored lots of interesting stuff online (e.g. http://boredzo.org/blog/archives/2008-01-20/asl-logging, https://github.com/robbiehanson/CocoaLumberjack) and the best hope seemed to be asl_open() with Facility of "com.apple.console" but alas, the output still doesn't show up in Console. Is NSLog the only option?
Add STDERR to ASL and then it shows up in the console
asl_add_log_file(NULL, STDERR_FILENO);
You need to set read privileges on the message, using either ReadUID or ReadGID. Setting either to -1 will allow any user/group to view the message according to the header file documentation.
aslmsg msg = asl_new(ASL_TYPE_MSG);
asl_set(msg, ASL_KEY_READ_UID, "-1");
asl_log(NULL, msg, ASL_LEVEL_NOTICE, "Hello, world!");
asl_free(msg);
I'm trying to debug RestKit object mapping and noticed that there are calls to RKLogDebug throughout the code, but it appears that that macro is undefined somewhere. How can I enable it?
You want to add something like this:
RKLogConfigureByName("RestKit", RKLogLevelWarning);
RKLogConfigureByName("RestKit/ObjectMapping", RKLogLevelTrace);
RKLogConfigureByName("RestKit/Network", RKLogLevelTrace);
to your code. See RKLog.h for the various levels. It is pretty trick.
N.B. this supports a wildcard at the end so, e.g.,
RKLogConfigureByName("*", RKLogLevelTrace); // set all logs to trace,
RKLogConfigureByName("RestKit*", RKLogLevelWarning); // set all RestKit logs to warning (leaving the app-specific log untouched).
– Thanks Kevin!
For Swift user use this syntex:
RKlcl_configure_by_name("RestKit/Network", RKlcl_vTrace.rawValue)
RKlcl_configure_by_na`enter code here`me("RestKit/ObjectMapping", RKlcl_vOff.rawValue)
– Thanks Darshit!
As described in first answer you can configure your app to specific component by calling RKLogConfigureByName.
You can also configure RestKit for specific component using Environment Variables in Xcode scheme. This is useful especially when you have your app building continuously for different environments.
Here's detailed explanation of RestKit logging http://restkit-tutorials.com/logging-in-restkit-debug-tips/
Are there any frameworks/services for logging user usage of an iphone application?
Say you want to log events like 'creating a contact' or something similiar, to know how people are actually using your app.
I have used and really like Pinch Media. (Recently renamed to flurry)
http://www.flurry.com
It gives you a bunch of analytics without you having to do anything special.
You can also add hooks to your code to see how often a particular section of code is run and get analytics on that area as well.
Yes, Flurry Analytics does this. You can even pass parameters to your events so you can break down which options are most popular, etc.
What you could do is simply log these messages perhaps as a Log custom class, and whenever you need to add anything to the log, you would simply add a new Log item to the AppDelegate or wherever you need to store the list of Logs. Then you can save them to the file system.
For achieving the logging the feature usage in flurry the url below can be helpful.
http://wiki.flurry.com/index.php?title=Analytics_-_Best_Practices
I was told this a few times in this very site, but I wanted to make sure this is really the case.
I was expecting to be able to sprinkle NSLog function calls throughout my code, and that Xcode/gcc would automatically strip those calls out when building my Release/Distribution builds.
Should I avoid using this? If so, what alternatives are most common between experienced Objective-C programmers?
Preprocessor macros are indeed great for debugging. There's nothing wrong with NSLog(), but it's simple to define your own logging function with better functionality. Here's the one I use, it includes the file name and line number to make it easier to track down log statements.
#define DEBUG_MODE
#ifdef DEBUG_MODE
#define DebugLog( s, ... ) NSLog( #"<%p %#:(%d)> %#", self, [[NSString stringWithUTF8String:__FILE__] lastPathComponent], __LINE__, [NSString stringWithFormat:(s), ##__VA_ARGS__] )
#else
#define DebugLog( s, ... )
#endif
I found it easier to put this entire statement in the prefix header rather than its own file. You could, if you wanted, build a more complicated logging system by having DebugLog interact with normal Objective-C objects. For instance, you could have a logging class that writes to its own log file (or database), and includes a 'priority' argument you could set at runtime, so debug messages are not shown in your release version, but error messages are (if you did this you could make DebugLog(), WarningLog(), and so on).
Oh, and keep in mind #define DEBUG_MODE can be re-used in different places in your application. For example, in my application I use it to disable license key checks and only allow the application to run if it's before a certain date. This lets me distribute a time limited, fully functional beta copy with minimal effort on my part.
Put this 3 lines at the end of -prefix.pch file:
#ifndef DEBUG
#define NSLog(...) /* suppress NSLog when in release mode */
#endif
You don't need to define anything into your project, because DEBUG is defined in your build setting by default when you create your project.
NSLog calls can be left in production code, but should only be there for truly exceptional cases, or information that it is desired that will be logged to the system log.
Applications which litter the system log are annoying, and come across as unprofessional.
I can't comment on Marc Charbonneau's answer, so I'll post this as an answer.
Further to adding the macro to your pre-compiled header, you can use the Target build configurations to control the defining (or lack of defining) the DEBUG_MODE.
If you select "Debug" active configuration, DEBUG_MODE will be defined, and the macro expands to the full NSLog definition.
Selecting the "Release" active configuration will not define DEBUG_MODE and your NSLogging is omitted from the release build.
Steps:
Target > Get Info
Build tab
Search for "PreProcessor Macros" (or GCC_PREPROCESSOR_DEFINITIONS)
Select Configuration: Debug
Edit Definition at this Level
Add DEBUG_MODE=1
Select Configuration: Release
confirm DEBUG_MODE is not set in GCC_PREPROCESSOR_DEFINITIONS
if you omit the '=' character in the definition, you will get an error from the preprocessor
Also, paste this comment (shown below) above the macro definition to remind you where the DEBUG_MACRO define comes from ;)
// Target > Get Info > Build > GCC_PREPROCESSOR_DEFINITIONS
// Configuration = Release: <empty>
// = Debug: DEBUG_MODE=1
EDIT: The method posted by Marc Charbonneau, and brought to my attention by sho, is far better than this one.
I have deleted the portion of my answer which suggested using an empty function to disable logging when debug mode is disabled. The portion that deals with setting an automatic preprocessor macro is still relevant, so it remains. I have also edited the name of the preprocessor macro so that it fits better with Marc Charbonneau's answer.
To achieve the automatic (and expected) behaviour in Xcode:
In the project settings, go to the "Build" tab, and select the "Debug" configuration. Find the "Preprocessor Macros" section, and add a macro named DEBUG_MODE.
...
EDIT: See Marc Charbonneau's answer for the proper way to enable and disable logging with the DEBUG_MODE macro.
I agree with Matthew. There's nothing wrong with NSLog in production code. In fact, it can be useful to the user. That said, if the only reason you're using NSLog is to help debug, then, yes, that should be removed before you release.
Furthermore, since you've tagged this as an iPhone question, NSLog takes resources, which is something the iPhone has precious little of. If you're NSLogging anything on the iPhone, that takes away processor time from your app. Use it wisely.
The simple truth is that NSLog is just plain slow.
But why? To answer that question, let's find out what NSLog does, and then how it does it.
What does NSLog do exactly?
NSLog does 2 things:
It writes log messages to the Apple System Logging (asl) facility. This allows log messages to show up in Console.app.
It also checks to see if the application's stderr stream is going to a terminal (such as when the application is being run via Xcode). If so it writes the log message to stderr (so that it shows up in the Xcode console).
Writing to STDERR doesn't sound difficult. That can be accomplished with fprintf and the stderr file descriptor reference. But what about asl?
The best documentation I've found about ASL is a 10 part blog post from Peter Hosey: link
Without going into too much detail, the highlight (as it concerns performance) is this:
To send a log message to the ASL facility, you basically open a client connection to the ASL daemon and send the message. BUT - each thread must use a separate client connection. So, to be thread safe, every time NSLog is called it opens a new asl client connection, sends the message, and then closes the connection.
Resources could be found here & here.
As noted in other answers you can use a #define to alter whether NSLog is used or not at compile time.
However a more flexible way is to use a logging library like Cocoa Lumberjack thatallows you to change whether something is logged at runtime as well.
In your code replace NSLog by DDLogVerbose or DDLogError etc, add a #import for the macro definitions etc and setup the loggers, often in the applicationDidFinishLaunching method.
To have the same effect as NSLog the configuration code is
[DDLog addLogger:[DDASLLogger sharedInstance]];
[DDLog addLogger:[DDTTYLogger sharedInstance]];
From a security point of view, it depends on what is being logged. If NSLog (or other loggers) is writing sensitive information, then you should remove the logger in production code.
From an auditing point of view, the auditor does not want to look at each use of NSLog to ensure its not logging sensitive information. He/she will simply tell you to remove the logger.
I work with both groups. We audit code, write the coding guides, etc. Our guide requires that logging is disabled in production code. So the internal teams know not to try it ;)
We will also reject an external app that logs in production because we don't want to accept the risk associated with accidentally leaking sensitive information. We don't care what the developer tells us. Its simply not worth our time to investigate.
And remember, we define 'sensitive', and not the developer ;)
I also perceive an app which performs lots of logging as an app ready to implode. There's a reason so much logging is performed/needed, and its usually not stability. Its right up there with 'watchdog' threads that restart hung services.
If you have never been through a Security Architecture (SecArch) review, these are the sorts of things we look at.
You shouldn't be needlessly verbose with printf or NSLog in release code. Try only doing a printf or NSLog if the app has something bad happen to it, I.E. an unrecoverable error.
Keep in mind that NSLogs can slow down the UI / main thread. It is best to remove them from release builds unless absolutely necessary.
I would highly recommend using TestFlight for logging (free). Their method will override NSLog (using a macro) and allow you to turn on/off logging to their server, Apple System log and STDERR log, for all your existing calls to NSLog. The nice thing about this is you can still review your log messages for apps deployed to testers and apps deployed in the App Store, without the logs appearing on the user's system log. The best of both worlds.
I have an AIR application that takes command-line arguments via onInvoke. All is good, but I cannot figure out how to print some status messages back to the user (to stdout / console, so to speak). Is it possible?
Even a default log file for traces would be fine, but I can't find any info about it anywhere. Do I need to create my own log file? Now that'd be silly.
Take a look at CommandProxy. It is a low level wrapper around your AIR application that lets you send command from AS3 back to the proxy for communicating with the underlying OS. You should be able to add a means of writing to the command line via such a method.
I don't think that is possible, but I'm not completely sure though.
There is a flashlog.txt file which you can configure so all trace() statements are logged to it. Check this post http://www.digitalflipbook.com/archives/2005/07/trace_from_the.php for more info on how to set it up. This is for logging from the browser, but I'm pretty sure it should also work from an air app.
Additionally, you could use SOS MAX from Powerflasher to log to an external console through an XML socket.
By default, trace() will output to stdout.
Your AIR application is one, big trace window if you want it to be.