third_party/java/proguard/proguard6.2.2/docs/manual/examples.md
You can find some sample configuration files in the examples directory
of the ProGuard distribution.
To shrink, optimize, and obfuscate a simple Java application, you
typically create a configuration file like myconfig.pro, which you can
then use with
bin/proguard @myconfig.pro
The configuration file specifies the input, the output, and the entry points of the application:
-injars myapplication.jar
-outjars myapplication_out.jar
-libraryjars <java.home>/jmods/java.base.jmod(!**.jar;!module-info.class)
-printmapping myapplication.map
-keep public class com.example.MyMain {
public static void main(java.lang.String[]);
}
Note the use of the <java.home> system property. ProGuard automatically
replaces it when parsing the file. In this example, the library jar is the
base Java runtime module, minus some unwanted files. For Java 8 or older, the
Java runtime jar would be <java.home>/lib/rt.jar instead. You may need
additional modules or jars if your application depends on them.
The -keep option specifies the entry point of the
application that has to be preserved. The access modifiers public and
static are not really required in this case, since we know a priori that the
specified class and method have the proper access flags. It just looks more
familiar this way.
Note that all type names are fully specified: com.example.MyMain and
java.lang.String[].
We're writing out an obfuscation mapping file with
-printmapping, for de-obfuscating any stack traces
later on, or for incremental obfuscation of extensions.
We can further improve the results with a few additional options:
-optimizationpasses 3
-overloadaggressively
-repackageclasses ''
-allowaccessmodification
These options are not required; they just shave off some extra bytes from the output jar, by performing up to 3 optimization passes, and by aggressively obfuscating class members and package names.
In general, you might need a few additional options for processing native methods, callback methods, enumerations, serializable classes, bean classes, annotations, and resource files.
These options shrink, optimize, and obfuscate the applet
com.example.MyApplet:
-injars in.jar
-outjars out.jar
-libraryjars <java.home>/jmods/java.base.jmod(!**.jar;!module-info.class)
-keep public class com.example.MyApplet
The typical applet methods will be preserved automatically, since
com.example.MyApplet is an extension of the Applet class in the library
rt.jar.
If applicable, you should add options for processing native methods, callback methods, enumerations, serializable classes, bean classes, annotations, and resource files.
These options shrink, optimize, obfuscate, and preverify the midlet
com.example.MyMIDlet:
-injars in.jar
-outjars out.jar
-libraryjars /usr/local/java/wtk2.5.2/lib/midpapi20.jar
-libraryjars /usr/local/java/wtk2.5.2/lib/cldcapi11.jar
-overloadaggressively
-repackageclasses ''
-allowaccessmodification
-microedition
-keep public class com.example.MyMIDlet
Note how we're now targeting the Java Micro Edition run-time environment of
midpapi20.jar and cldcapi11.jar, instead of the Java Standard Edition
run-time environment rt.jar. You can target other JME environments by
picking the appropriate jars.
The typical midlet methods will be preserved automatically, since
com.example.MyMIDlet is an extension of the MIDlet class in the library
midpapi20.jar.
The -microedition option makes sure the class files
are preverified for Java Micro Edition, producing compact StackMap
attributes. It is no longer necessary to run an external preverifier.
Be careful if you do use the external preverify tool on a platform with a
case-insensitive filing system, such as Windows. Because this tool unpacks
your processed jars, you should then use ProGuard's
-dontusemixedcaseclassnames option.
If applicable, you should add options for processing native methods and resource files.
Note that you will still have to adapt the midlet jar size in the corresponding jad file; ProGuard doesn't do that for you.
These options shrink, optimize, and obfuscate the Java Card applet
com.example.MyApplet:
-injars in.jar
-outjars out.jar
-libraryjars /usr/local/java/javacard2.2.2/lib/api.jar
-dontwarn java.lang.Class
-overloadaggressively
-repackageclasses ''
-allowaccessmodification
-keep public class com.example.MyApplet
The configuration is very similar to the configuration for midlets, except that it now targets the Java Card run-time environment. This environment doesn't have java.lang.Class, so we're telling ProGuard not to worry about it.
These options shrink, optimize, and obfuscate the xlet com.example.MyXlet:
-injars in.jar
-outjars out.jar
-libraryjars /usr/local/java/jtv1.1/javatv.jar
-libraryjars /usr/local/java/cdc1.1/lib/cdc.jar
-libraryjars /usr/local/java/cdc1.1/lib/btclasses.zip
-overloadaggressively
-repackageclasses ''
-allowaccessmodification
-keep public class com.example.MyXlet
The configuration is very similar to the configuration for midlets, except that it now targets the CDC run-time environment with the Java TV API.
These options shrink, optimize, and obfuscate the single Android activity
com.example.MyActivity:
-injars bin/classes
-outjars bin/classes-processed.jar
-libraryjars /usr/local/java/android-sdk/platforms/android-9/android.jar
-android
-dontpreverify
-repackageclasses ''
-allowaccessmodification
-optimizations !code/simplification/arithmetic
-keep public class com.example.MyActivity
We're targeting the Android run-time and keeping the activity as an entry point.
Preverification is irrelevant for the dex compiler and the Dalvik VM, so we
can switch it off with the -dontpreverify option.
The -optimizations option disables some arithmetic
simplifications that Dalvik 1.0 and 1.5 can't handle. Note that the Dalvik VM
also can't handle aggressive overloading (of
static fields).
If applicable, you should add options for processing native methods, callback methods, enumerations, annotations, and resource files.
!!! note ""
{: .icon} The standard build processes of the
Android SDK (with Ant, Gradle, Android Studio, and Eclipse) already
integrate ProGuard with all the proper settings. You only need to enable
ProGuard by uncommenting the line "proguard.config=....." in the file
project.properties of your Ant project, or by adapting the
build.gradle file of your Gradle project. You then don't need any of
the configuration below.
Notes:
proguard-project.txt,...) contain the necessary
settings for your application.${sdk.dir}/tools/proguard/proguard-android-optimize.txt instead of the
default ${sdk.dir}/tools/proguard/proguard-android.txt.For more information, you can consult the official Developer Guide in the Android SDK.
If you're constructing a build process from scratch: these options shrink, optimize, and obfuscate all public activities, services, broadcast receivers, and content providers from the compiled classes and external libraries:
-injars bin/classes
-injars bin/resources.ap_
-injars libs
-outjars bin/application.apk
-libraryjars /usr/local/android-sdk/platforms/android-28/android.jar
-android
-dontpreverify
-repackageclasses ''
-allowaccessmodification
-optimizations !code/simplification/arithmetic
-keepattributes *Annotation*
-keep public class * extends android.app.Activity
-keep public class * extends android.app.Application
-keep public class * extends android.app.Service
-keep public class * extends android.content.BroadcastReceiver
-keep public class * extends android.content.ContentProvider
-keep public class * extends android.view.View {
public <init>(android.content.Context);
public <init>(android.content.Context, android.util.AttributeSet);
public <init>(android.content.Context, android.util.AttributeSet, int);
public void set*(...);
}
-keepclasseswithmembers class * {
public <init>(android.content.Context, android.util.AttributeSet);
}
-keepclasseswithmembers class * {
public <init>(android.content.Context, android.util.AttributeSet, int);
}
-keepclassmembers class * extends android.content.Context {
public void *(android.view.View);
public void *(android.view.MenuItem);
}
-keepclassmembers class * implements android.os.Parcelable {
static ** CREATOR;
}
-keepclassmembers class **.R$* {
public static <fields>;
}
-keepclassmembers class * {
@android.webkit.JavascriptInterface <methods>;
}
Most importantly, we're keeping all fundamental classes that may be referenced
by the AndroidManifest.xml file of the application. If your manifest file
contains other classes and methods, you may have to specify those as well.
We're keeping annotations, since they might be used by custom RemoteViews
and by various frameworks.
We're keeping any custom View extensions and other classes with typical
constructors, since they might be referenced from XML layout files.
We're also keeping possible onClick handlers in custom Context extensions,
since they might be referenced from XML layout files.
We're also keeping the required static fields in Parcelable implementations,
since they are accessed by introspection.
We're keeping the static fields of referenced inner classes of auto-generated
R classes, just in case your code is accessing those fields by
introspection. Note that the compiler already inlines primitive fields, so
ProGuard can generally remove all these classes entirely anyway (because the
classes are not referenced and therefore not required).
Finally, we're keeping annotated Javascript interface methods, so they can be exported and accessed by their original names. Javascript interface methods that are not annotated (in code targeted at Android versions older than 4.2) still need to be preserved manually.
If you're using additional Google APIs, you'll have to specify those as well, for instance:
-libraryjars /usr/local/java/android-sdk/extras/android/support/v4/android-support-v4.jar
-libraryjars /usr/local/java/android-sdk/add-ons/addon-google_apis-google-21/libs/maps.jar
If you're using Google's optional License Verification Library, you can
obfuscate its code along with your own code. You do have to preserve its
ILicensingService interface for the library to work:
-keep public interface com.android.vending.licensing.ILicensingService
If you're using the Android Compatibility library, you should add the following line, to let ProGuard know it's ok that the library references some classes that are not available in all versions of the API:
-dontwarn android.support.**
If applicable, you should add options for processing native
methods, callback methods,
enumerations, and resource files. You may
also want to add options for producing useful stack traces and
to remove logging. You can find a complete sample configuration in
examples/standalone/android.pro in the ProGuard distribution.
These options shrink, optimize, and obfuscate an entire library, keeping all public and protected classes and class members, native method names, and serialization code. The processed version of the library can then still be used as such, for developing code based on its public API.
-injars in.jar
-outjars out.jar
-libraryjars <java.home>/jmods/java.base.jmod(!**.jar;!module-info.class)
-printmapping out.map
-keep public class * {
public protected *;
}
-keepparameternames
-renamesourcefileattribute SourceFile
-keepattributes Exceptions,InnerClasses,Signature,Deprecated,
SourceFile,LineNumberTable,*Annotation*,EnclosingMethod
-keepclasseswithmembernames,includedescriptorclasses class * {
native <methods>;
}
-keepclassmembers,allowoptimization enum * {
public static **[] values();
public static ** valueOf(java.lang.String);
}
-keepclassmembers class * implements java.io.Serializable {
static final long serialVersionUID;
private static final java.io.ObjectStreamField[] serialPersistentFields;
private void writeObject(java.io.ObjectOutputStream);
private void readObject(java.io.ObjectInputStream);
java.lang.Object writeReplace();
java.lang.Object readResolve();
}
This configuration should preserve everything a developers ever wants to
access in the library. Only if there are any other non-public classes or
methods that are invoked dynamically, they should be specified using
additional -keep options.
The "Exceptions" attribute has to be preserved, so the compiler knows which exceptions methods may throw.
The "InnerClasses" attribute (or more precisely, its source name part) has to
be preserved too, for any inner classes that can be referenced from outside
the library. The javac compiler would be unable to find the inner classes
otherwise.
The "Signature" attribute is required to be able to access generic types when compiling in JDK 5.0 and higher.
The -keepparameternames option keeps the
parameter names in the "LocalVariableTable" and "LocalVariableTypeTable"
attributes of public library methods. Some IDEs can present these names to the
developers who use the library.
Finally, we're keeping the "Deprecated" attribute and the attributes for producing useful stack traces.
We've also added some options for for processing native methods, enumerations, serializable classes, and annotations, which are all discussed in their respective examples.
These options shrink, optimize, and obfuscate all public applications in
in.jar:
-injars in.jar
-outjars out.jar
-libraryjars <java.home>/jmods/java.base.jmod(!**.jar;!module-info.class)
-printseeds
-keepclasseswithmembers public class * {
public static void main(java.lang.String[]);
}
Note the use of -keepclasseswithmembers.
We don't want to preserve all classes, just all classes that have main
methods, and those methods.
The -printseeds option prints out which classes
exactly will be preserved, so we know for sure we're getting what we want.
If applicable, you should add options for processing native methods, callback methods, enumerations, serializable classes, bean classes, annotations, and resource files.
These options shrink, optimize, and obfuscate all public applets in in.jar:
-injars in.jar
-outjars out.jar
-libraryjars <java.home>/jmods/java.base.jmod(!**.jar;!module-info.class)
-libraryjars <java.home>/jmods/java.desktop.jmod(!**.jar;!module-info.class)
-printseeds
-keep public class * extends java.applet.Applet
We're simply keeping all classes that extend the Applet class.
Again, the -printseeds option prints out which
applets exactly will be preserved.
If applicable, you should add options for processing native methods, callback methods, enumerations, serializable classes, bean classes, annotations, and resource files.
These options shrink, optimize, obfuscate, and preverify all public midlets in
in.jar:
-injars in.jar
-outjars out.jar
-libraryjars /usr/local/java/wtk2.5.2/lib/midpapi20.jar
-libraryjars /usr/local/java/wtk2.5.2/lib/cldcapi11.jar
-overloadaggressively
-repackageclasses ''
-allowaccessmodification
-microedition
-printseeds
-keep public class * extends javax.microedition.midlet.MIDlet
We're simply keeping all classes that extend the MIDlet class.
The -microedition option makes sure the class files
are preverified for Java Micro Edition, producing compact StackMap
attributes. It is no longer necessary to run an external preverifier.
Be careful if you do use the external preverify tool on a platform with a
case-insensitive filing system, such as Windows. Because this tool unpacks
your processed jars, you should then use ProGuard's
-dontusemixedcaseclassnames option.
The -printseeds option prints out which midlets
exactly will be preserved.
If applicable, you should add options for processing native methods and resource files.
Note that you will still have to adapt the midlet jar size in the corresponding jad file; ProGuard doesn't do that for you.
These options shrink, optimize, and obfuscate all public Java Card
applets in in.jar:
-injars in.jar
-outjars out.jar
-libraryjars /usr/local/java/javacard2.2.2/lib/api.jar
-dontwarn java.lang.Class
-overloadaggressively
-repackageclasses ''
-allowaccessmodification
-printseeds
-keep public class * implements javacard.framework.Applet
We're simply keeping all classes that implement the Applet interface.
The -printseeds option prints out which applets
exactly will be preserved.
These options shrink, optimize, and obfuscate all public xlets in in.jar:
-injars in.jar
-outjars out.jar
-libraryjars /usr/local/java/jtv1.1/javatv.jar
-libraryjars /usr/local/java/cdc1.1/lib/cdc.jar
-libraryjars /usr/local/java/cdc1.1/lib/btclasses.zip
-overloadaggressively
-repackageclasses ''
-allowaccessmodification
-printseeds
-keep public class * implements javax.tv.xlet.Xlet
We're simply keeping all classes that implement the Xlet interface.
The -printseeds option prints out which xlets exactly
will be preserved.
These options shrink, optimize, and obfuscate all public servlets in in.jar:
-injars in.jar
-outjars out.jar
-libraryjars <java.home>/lib/rt.jar
-libraryjars /usr/local/java/servlet/servlet.jar
-printseeds
-keep public class * implements javax.servlet.Servlet
Keeping all servlets is very similar to keeping all applets. The servlet API is not part of the standard run-time jar, so we're specifying it as a library. Don't forget to use the right path name.
We're then keeping all classes that implement the Servlet interface. We're
using the implements keyword because it looks more familiar in this context,
but it is equivalent to extends, as far as ProGuard is concerned.
The -printseeds option prints out which
servlets exactly will be preserved.
If applicable, you should add options for processing native methods, callback methods, enumerations, serializable classes, bean classes, annotations, and resource files.
These options shrink, optimize, and obfuscate all public Scala applications in
in.jar:
-injars in.jar
-injars /usr/local/java/scala-2.9.1/lib/scala-library.jar
-outjars out.jar
-libraryjars <java.home>/jmods/java.base.jmod(!**.jar;!module-info.class)
-dontwarn scala.**
-keepclasseswithmembers public class * {
public static void main(java.lang.String[]);
}
-keep class * implements org.xml.sax.EntityResolver
-keepclassmembers class * {
** MODULE$;
}
-keepclassmembernames class scala.concurrent.forkjoin.ForkJoinPool {
long eventCount;
int workerCounts;
int runControl;
scala.concurrent.forkjoin.ForkJoinPool$WaitQueueNode syncStack;
scala.concurrent.forkjoin.ForkJoinPool$WaitQueueNode spareStack;
}
-keepclassmembernames class scala.concurrent.forkjoin.ForkJoinWorkerThread {
int base;
int sp;
int runState;
}
-keepclassmembernames class scala.concurrent.forkjoin.ForkJoinTask {
int status;
}
-keepclassmembernames class scala.concurrent.forkjoin.LinkedTransferQueue {
scala.concurrent.forkjoin.LinkedTransferQueue$PaddedAtomicReference head;
scala.concurrent.forkjoin.LinkedTransferQueue$PaddedAtomicReference tail;
scala.concurrent.forkjoin.LinkedTransferQueue$PaddedAtomicReference cleanMe;
}
The configuration is essentially the same as for processing applications, because Scala is compiled to ordinary Java bytecode. However, the example processes the Scala runtime library as well. The processed jar can be an order of magnitude smaller and a few times faster than the original code (for the Scala code examples, for instance).
The -dontwarn option tells ProGuard not to complain
about some artefacts in the Scala runtime, the way it is compiled by the
scalac compiler (at least in Scala 2.9.1 and older). Note that this option
should always be used with care.
The additional -keep options make sure that some
classes and some fields that are accessed by means of introspection are not
removed or renamed.
If applicable, you should add options for processing native methods, callback methods, enumerations, serializable classes, bean classes, annotations, and resource files.
If your application, applet, servlet, library, etc., contains native methods, you'll want to preserve their names and their classes' names, so they can still be linked to the native library. The following additional option will ensure that:
-keepclasseswithmembernames,includedescriptorclasses class * {
native <methods>;
}
Note the use of
-keepclasseswithmembernames. We don't
want to preserve all classes or all native methods; we just want to keep the
relevant names from being obfuscated. The modifier
includedescriptorclasses additionally
makes sure that the return types and parameter types aren't renamed either, so
the entire signatures remain compatible with the native libraries.
ProGuard doesn't look at your native code, so it won't automatically preserve the classes or class members that are invoked by the native code. These are entry points, which you'll have to specify explicitly. Callback methods are discussed below as a typical example.
If your application, applet, servlet, library, etc., contains callback
methods, which are called from external code (native code, scripts,...),
you'll want to preserve them, and probably their classes too. They are just
entry points to your code, much like, say, the main method of an application.
If they aren't preserved by other -keep options, something
like the following option will keep the callback class and method:
-keep class com.example.MyCallbackClass {
void myCallbackMethod(java.lang.String);
}
This will preserve the given class and method from being removed or renamed.
If your application, applet, servlet, library, etc., contains enumeration classes, you'll have to preserve some special methods. Enumerations were introduced in Java 5. The java compiler translates enumerations into classes with a special structure. Notably, the classes contain implementations of some static methods that the run-time environment accesses by introspection (Isn't that just grand? Introspection is the self-modifying code of a new generation). You have to specify these explicitly, to make sure they aren't removed or obfuscated:
-keepclassmembers,allowoptimization enum * {
public static **[] values();
public static ** valueOf(java.lang.String);
}
More complex applications, applets, servlets, libraries, etc., may contain classes that are serialized. Depending on the way in which they are used, they may require special attention:
Often, serialization is simply a means of transporting data, without long-term storage. Classes that are shrunk and obfuscated should then continue to function fine with the following additional options:
-keepclassmembers class * implements java.io.Serializable {
private static final java.io.ObjectStreamField[] serialPersistentFields;
private void writeObject(java.io.ObjectOutputStream);
private void readObject(java.io.ObjectInputStream);
java.lang.Object writeReplace();
java.lang.Object readResolve();
}
The -keepclassmembers option makes sure
that any serialization methods are kept. By using this option instead of
the basic -keep option, we're not forcing preservation of all
serializable classes, just preservation of the listed members of classes
that are actually used.
Sometimes, the serialized data are stored, and read back later into newer
versions of the serializable classes. One then has to take care the classes
remain compatible with their unprocessed versions and with future processed
versions. In such cases, the relevant classes will most likely have
serialVersionUID fields. The following options should then be sufficient
to ensure compatibility over time:
-keepnames class * implements java.io.Serializable
-keepclassmembers class * implements java.io.Serializable {
static final long serialVersionUID;
private static final java.io.ObjectStreamField[] serialPersistentFields;
!static !transient <fields>;
private void writeObject(java.io.ObjectOutputStream);
private void readObject(java.io.ObjectInputStream);
java.lang.Object writeReplace();
java.lang.Object readResolve();
}
The serialVersionUID and serialPersistentFields lines makes sure those
fields are preserved, if they are present. The <fields> line preserves
all non-static, non-transient fields, with their original names. The
introspection of the serialization process and the de-serialization
process will then find consistent names.
Occasionally, the serialized data have to remain compatible, but the classes
involved lack serialVersionUID fields. I imagine the original code will
then be hard to maintain, since the serial version UID is then computed from
a list of features the serializable class. Changing the class ever so
slightly may change the computed serial version UID. The list of features is
specified in the section on Stream Unique
Identifiers
of Sun's Java Object Serialization
Specification.
The following directives should at least partially ensure compatibility with
the original classes:
-keepnames class * implements java.io.Serializable
-keepclassmembers class * implements java.io.Serializable {
static final long serialVersionUID;
private static final java.io.ObjectStreamField[] serialPersistentFields;
!static !transient <fields>;
!private <fields>;
!private <methods>;
private void writeObject(java.io.ObjectOutputStream);
private void readObject(java.io.ObjectInputStream);
java.lang.Object writeReplace();
java.lang.Object readResolve();
}
The new options force preservation of the elements involved in the UID
computation. In addition, the user will have to manually specify all
interfaces of the serializable classes (using something like "-keep interface MyInterface"), since these names are also used when computing
the UID. A fast but sub-optimal alternative would be simply keeping all
interfaces with "-keep interface *".
In the rare event that you are serializing lambda expressions in Java 8 or higher, you need to preserve some methods and adapt the hard-coded names of the classes in which they occur:
-keepclassmembers class * {
private static synthetic java.lang.Object $deserializeLambda$(java.lang.invoke.SerializedLambda);
}
-keepclassmembernames class * {
private static synthetic *** lambda$*(...);
}
-adaptclassstrings com.example.Test
This should satisfy the reflection in the deserialization code of the Java run-time.
Note that the above options may preserve more classes and class members than
strictly necessary. For instance, a large number of classes may implement the
Serialization interface, yet only a small number may actually ever be
serialized. Knowing your application and tuning the configuration often
produces more compact results.
If your application, applet, servlet, library, etc., makes extensive use of introspection on bean classes to find bean editor classes, or getter and setter methods, then configuration may become painful. There's not much else you can do than making sure the bean class names, or the getter and setter names don't change. For instance:
-keep public class com.example.MyBean {
public void setMyProperty(int);
public int getMyProperty();
}
-keep public class com.example.MyBeanEditor
If there are too many elements to list explicitly, wildcards in class
names and method signatures might be helpful. This example preserves all
possible setters and getters in classes in the package mybeans:
-keep class mybeans.** {
void set*(***);
void set*(int, ***);
boolean is*();
boolean is*(int);
*** get*();
*** get*(int);
}
The '***' wildcard matches any type (primitive or non-primitive, array
or non-array). The methods with the 'int' arguments matches properties
that are lists.
If your application, applet, servlet, library, etc., uses annotations, you may want to preserve them in the processed output. Annotations are represented by attributes that have no direct effect on the execution of the code. However, their values can be retrieved through introspection, allowing developers to adapt the execution behavior accordingly. By default, ProGuard treats annotation attributes as optional, and removes them in the obfuscation step. If they are required, you'll have to specify this explicitly:
-keepattributes *Annotation*
For brevity, we're specifying a wildcarded attribute name, which will match
RuntimeVisibleAnnotations, RuntimeInvisibleAnnotations,
RuntimeVisibleParameterAnnotations, RuntimeInvisibleParameterAnnotations,
and AnnotationDefault. Depending on the purpose of the processed code, you
could refine this selection, for instance not keeping the run-time invisible
annotations (which are only used at compile-time).
Some code may make further use of introspection to figure out the enclosing methods of anonymous inner classes. In that case, the corresponding attribute has to be preserved as well:
-keepattributes EnclosingMethod
Database drivers are implementations of the Driver interface. Since they are
often created dynamically, you may want to preserve any implementations that
you are processing as entry points:
-keep class * implements java.sql.Driver
This option also gets rid of the note that ProGuard prints out about
(java.sql.Driver)Class.forName constructs, if you are instantiating a driver
in your code (without necessarily implementing any drivers yourself).
Swing UI look and feels are implemented as extensions of the ComponentUI
class. For some reason, these have to contain a static method createUI,
which the Swing API invokes using introspection. You should therefore always
preserve the method as an entry point, for instance like this:
-keep class * extends javax.swing.plaf.ComponentUI {
public static javax.swing.plaf.ComponentUI createUI(javax.swing.JComponent);
}
This option also keeps the classes themselves.
Reportedly, the easiest way to handle RMI code is to process the code with
ProGuard first and then invoke the rmic tool. If that is not possible, you
may want to try something like this:
-keepattributes Exceptions
-keep interface * extends java.rmi.Remote {
<methods>;
}
-keep class * implements java.rmi.Remote {
<init>(java.rmi.activation.ActivationID, java.rmi.MarshalledObject); {: #activation}
}
The first -keep option keeps all your Remote interfaces and
their methods. The second one keeps all the implementations, along with their
particular RMI constructors, if any.
The Exceptions attribute has to be kept too, because the RMI handling code
performs introspection to check whether the method signatures are compatible.
ProGuard optimizes Gson code, by detecting which domain classes are serialized using the Gson library, and then replacing the reflection-based implementation by more efficient hard-coded serialization.
The GSON optimization is enabled by default and doesn't require any additional configuration. If you've disabled optimization, the GSON library still relies on reflection on the fields of the classes that it serializes. You then need to preserve the parameterless constructor and the serialized fields from being removed, optimized, or obfuscated. For example:
-keepclassmembers class com.example.SerializedClass {
<fields>;
<init>();
}
While creating the configuration, you can specify the option
-addconfigurationdebugging, to get
feedback on the necessary settings at run-time.
Alternatively, you can make sure the fields are explicitly annotated with
@SerializedName, so the names of the fields can be obfuscated. You can
then keep all of them at the same time with:
-keepclasseswithmembers,allowobfuscation,includedescriptorclasses class * {
@com.google.gson.annotations.SerializedName <fields>;
}
-keepclassmembers enum * {
@com.google.gson.annotations.SerializedName <fields>;
}
If your application is using JEE-style dependency injection, the application
container will automatically assign instances of resource classes to fields
and methods that are annotated with @Resource. The container applies
introspection, even accessing private class members directly. It typically
constructs a resource name based on the type name and the class member name.
We then have to avoid that such class members are removed or renamed:
-keepclassmembers class * {
@javax.annotation.Resource *;
}
The Spring framework has another similar annotation @Autowired:
-keepclassmembers class * {
@org.springframework.beans.factory.annotation.Autowired *;
}
Your Android application may be using the Dagger library for its dependency injection.
Dagger 1 relies heavily on reflection, so you may need some additional configuration to make sure it continues to work. DexGuard's default configuration already keeps some required classes:
-keepclassmembers,allowobfuscation class * {
@dagger.** *;
}
-keep class **$$ModuleAdapter
-keep class **$$InjectAdapter
-keep class **$$StaticInjection
-if class **$$ModuleAdapter
-keep class <1>
-if class **$$InjectAdapter
-keep class <1>
-if class **$$StaticInjection
-keep class <1>
-keepnames class dagger.Lazy
That way, Dagger can combine the corresponding pairs of classes, based on their names.
Furthermore, if your code injects dependencies into some given classes with an
annotation like @Module(injects = { SomeClass.class }, ...), you need to
preserve the specified names as well:
-keep class com.example.SomeClass
Dagger 2 no longer relies on reflection. You don't need to preserve any classes there.
If your Android application includes Butterknife to inject views, you also need a few lines of configuration, since Butterknife relies on reflection to tie together the code at runtime:
-keep @interface butterknife.*
-keepclasseswithmembers class * {
@butterknife.* <fields>;
}
-keepclasseswithmembers class * {
@butterknife.* <methods>;
}
-keepclasseswithmembers class * {
@butterknife.On* <methods>;
}
-keep class **$$ViewInjector {
public static void inject(...);
public static void reset(...);
}
-keep class **$$ViewBinder {
public static void bind(...);
public static void unbind(...);
}
-if class **$$ViewBinder
-keep class <1>
-keep class **_ViewBinding {
<init>(<1>, android.view.View);
}
-if class **_ViewBinding
-keep class <1>
These settings preserve the Butterknife annotations, the annotated fields and methods, and the generated classes and methods that Butterknife accesses by reflection.
If your application, applet, servlet, library, etc., contains resource files, it may be necessary to adapt their names and/or their contents when the application is obfuscated. The following two options can achieve this automatically:
-adaptresourcefilenames **.properties,**.gif,**.jpg
-adaptresourcefilecontents **.properties,META-INF/MANIFEST.MF
The -adaptresourcefilenames option in
this case renames properties files and image files in the processed output,
based on the obfuscated names of their corresponding class files (if any). The
-adaptresourcefilecontents option
looks for class names in properties files and in the manifest file, and
replaces these names by the obfuscated names (if any). You'll probably want to
adapt the filters to suit your application.
As illustrated in the previous section, manifest files can be treated like ordinary resource files. ProGuard can adapt obfuscated class names in the files, but it won't make any other changes. If you want anything else, you should apply an external tool. For instance, if a manifest file contains signing information, you should sign the jar again after it has been processed.
If you're merging several input jars into a single output jar, you'll have to pick one, typically by specifying filters:
-injars in1.jar
-injars in2.jar(!META-INF/MANIFEST.MF)
-injars in3.jar(!META-INF/MANIFEST.MF)
-outjars out.jar
The filters will let ProGuard copy the manifest file from the first jar and ignore any manifest files in the second and third input jars. Note that ProGuard will leave the order of the files in the jars unchanged; manifest files are not necessarily put first.
These options let obfuscated applications or libraries produce stack traces that can still be deciphered later on:
-printmapping out.map
-renamesourcefileattribute SourceFile
-keepattributes SourceFile,LineNumberTable
We're keeping all source file attributes, but we're replacing their values by the string "SourceFile". We could use any string. This string is already present in all class files, so it doesn't take up any extra space. If you're working with J++, you'll want to keep the "SourceDir" attribute as well.
We're also keeping the line number tables of all methods.
Whenever both of these attributes are present, the Java run-time environment will include line number information when printing out exception stack traces.
The information will only be useful if we can map the obfuscated names back to
their original names, so we're saving the mapping to a file out.map. The
information can then be used by the ReTrace tool to
restore the original stack trace.
Package names can be obfuscated in various ways, with increasing levels of obfuscation and compactness. For example, consider the following classes:
mycompany.myapplication.MyMain
mycompany.myapplication.Foo
mycompany.myapplication.Bar
mycompany.myapplication.extra.FirstExtra
mycompany.myapplication.extra.SecondExtra
mycompany.util.FirstUtil
mycompany.util.SecondUtil
Let's assume the class name mycompany.myapplication.MyMain is the main
application class that is kept by the configuration. All other class names can
be obfuscated.
By default, packages that contain classes that can't be renamed aren't renamed either, and the package hierarchy is preserved. This results in obfuscated class names like these:
mycompany.myapplication.MyMain
mycompany.myapplication.a
mycompany.myapplication.b
mycompany.myapplication.a.a
mycompany.myapplication.a.b
mycompany.a.a
mycompany.a.b
The -flattenpackagehierarchy option
obfuscates the package names further, by flattening the package hierarchy of
obfuscated packages:
-flattenpackagehierarchy 'myobfuscated'
The obfuscated class names then look as follows:
mycompany.myapplication.MyMain
mycompany.myapplication.a
mycompany.myapplication.b
myobfuscated.a.a
myobfuscated.a.b
myobfuscated.b.a
myobfuscated.b.b
Alternatively, the -repackageclasses option
obfuscates the entire packaging, by combining obfuscated classes into a single
package:
-repackageclasses 'myobfuscated'
The obfuscated class names then look as follows:
mycompany.myapplication.MyMain
mycompany.myapplication.a
mycompany.myapplication.b
myobfuscated.a
myobfuscated.b
myobfuscated.c
myobfuscated.d
Additionally specifying the
-allowaccessmodification option allows
access permissions of classes and class members to be broadened, opening up
the opportunity to repackage all obfuscated classes:
-repackageclasses 'myobfuscated'
-allowaccessmodification
The obfuscated class names then look as follows:
mycompany.myapplication.MyMain
myobfuscated.a
myobfuscated.b
myobfuscated.c
myobfuscated.d
myobfuscated.e
myobfuscated.f
The specified target package can always be the root package. For instance:
-repackageclasses ''
-allowaccessmodification
The obfuscated class names are then the shortest possible names:
mycompany.myapplication.MyMain
a
b
c
d
e
f
Note that not all levels of obfuscation of package names may be acceptable for all code. Notably, you may have to take into account that your application may contain resource files that have to be adapted.
You can let ProGuard remove logging code. The trick is to specify that the logging methods don't have side-effects — even though they actually do, since they write to the console or to a log file. ProGuard will take your word for it and remove the invocations (in the optimization step) and if possible the logging classes and methods themselves (in the shrinking step).
For example, this configuration removes invocations of the Android logging methods:
-assumenosideeffects class android.util.Log {
public static boolean isLoggable(java.lang.String, int);
public static int v(...);
public static int i(...);
public static int w(...);
public static int d(...);
public static int e(...);
}
The wildcards are a shortcut to match all versions of the methods. Be careful
not to use a * wildcard to match all methods, because it would also match
methods like wait(), higher up the hierarchy. Removing those invocations
will generally break your code.
Note that you generally can't remove logging code that uses
System.out.println, since you would be removing all invocations of
java.io.PrintStream#println, which could break your application. You can
work around it by creating your own logging methods and let ProGuard remove
those.
Logging statements often contain implicit calls that perform string concatenation. They no longer serve a purpose after the logging calls have been removed. You can let ProGuard clean up such constructs as well by providing additional hints:
-assumenoexternalsideeffects class java.lang.StringBuilder {
public java.lang.StringBuilder();
public java.lang.StringBuilder(int);
public java.lang.StringBuilder(java.lang.String);
public java.lang.StringBuilder append(java.lang.Object);
public java.lang.StringBuilder append(java.lang.String);
public java.lang.StringBuilder append(java.lang.StringBuffer);
public java.lang.StringBuilder append(char[]);
public java.lang.StringBuilder append(char[], int, int);
public java.lang.StringBuilder append(boolean);
public java.lang.StringBuilder append(char);
public java.lang.StringBuilder append(int);
public java.lang.StringBuilder append(long);
public java.lang.StringBuilder append(float);
public java.lang.StringBuilder append(double);
public java.lang.String toString();
}
-assumenoexternalreturnvalues public final class java.lang.StringBuilder {
public java.lang.StringBuilder append(java.lang.Object);
public java.lang.StringBuilder append(java.lang.String);
public java.lang.StringBuilder append(java.lang.StringBuffer);
public java.lang.StringBuilder append(char[]);
public java.lang.StringBuilder append(char[], int, int);
public java.lang.StringBuilder append(boolean);
public java.lang.StringBuilder append(char);
public java.lang.StringBuilder append(int);
public java.lang.StringBuilder append(long);
public java.lang.StringBuilder append(float);
public java.lang.StringBuilder append(double);
}
Be careful specifying your own assumptions, since they can easily break your code.
You can let ProGuard optimize the code for the range of Android versions that you intend to support — the range between the minimum SDK version and the maximum SDK version in your Android manifest. It then removes all code for SDK versions that are not relevant, for example in the various Android support libraries.
For example, if the minimum SDK version in your Android manifest is 19, you can optimize the code accordingly:
-assumevalues class android.os.Build$VERSION {
int SDK_INT = 19..2147483647;
}
You can also specify return values for methods. The "=" keyword and the
"return" keyword are equivalent. Be careful specifying assumptions, since
they can easily break your code.
In simple applications, all output classes and resources files are merged into a single jar. For example:
-injars classes
-injars in1.jar
-injars in2.jar
-injars in3.jar
-outjars out.jar
This configuration merges the processed versions of the files in the classes
directory and the three jars into a single output jar out.jar.
If you want to preserve the structure of your input jars (and/or apks, aars, wars, ears, jmods, zips, or directories), you can specify an output directory (or an apk, an aar, a war, an ear, a jmod, or a zip). For example:
-injars in1.jar
-injars in2.jar
-injars in3.jar
-outjars out
The input jars will then be reconstructed in the directory out, with their
original names.
You can also combine archives into higher level archives. For example:
-injars in1.jar
-injars in2.jar
-injars in3.jar
-outjars out.war
The other way around, you can flatten the archives inside higher level archives into simple archives:
-injars in.war
-outjars out.jar
This configuration puts the processed contents of all jars inside in.war
(plus any other contents of in.war) into out.jar.
If you want to combine input jars (and/or apks, aars, wars, ears, jmods, zips,
or directories) into output jars (and/or apks, aars, wars, ears, jmods, zips,
or directories), you can group the -injars and
-outjars options. For example:
-injars base_in1.jar
-injars base_in2.jar
-injars base_in3.jar
-outjars base_out.jar
-injars extra_in.jar
-outjars extra_out.jar
This configuration puts the processed results of all base_in*.jar jars into
base_out.jar, and the processed results of the extra_in.jar into
extra_out.jar. Note that only the order of the options matters; the
additional whitespace is just for clarity.
This grouping, archiving, and flattening can be arbitrarily complex. ProGuard always tries to package output archives in a sensible way, reconstructing the input entries as much as required.
If you want even greater control, you can add filters to the input and the output, filtering out apks, jars, aars, wars, ears, jmods, zips, and/or ordinary files. For example, if you want to disregard certain files from an input jar:
-injars in.jar(!images/**)
-outjars out.jar
This configuration removes any files in the images directory and its
subdirectories.
Such filters can be convenient for avoiding warnings about duplicate files in the output. For example, only keeping the manifest file from a first input jar:
-injars in1.jar
-injars in2.jar(!META-INF/MANIFEST.MF)
-injars in3.jar(!META-INF/MANIFEST.MF)
-outjars out.jar
Another useful application is ignoring unwanted files from the runtime library module:
-libraryjars <java.home>/jmods/java.base.jmod(!**.jar;!module-info.class)
The filter makes ProGuard disregard redundant jars inside the module, and module info classes that would only cause conflicts with duplicate names.
It is also possible to filter the jars (and/or apks, aabs, aars, wars, ears, jmods, zips) themselves, based on their names. For example:
-injars in(**/acme_*.jar;)
-outjars out.jar
Note the semi-colon in the filter; the filter in front of it applies to jar
names. In this case, only acme_*.jar jars are read from the directory in
and its subdirectories. Filters for war names, ear names, and zip names can be
prefixed with additional semi-colons. All types of filters can be combined.
They are orthogonal.
On the other hand, you can also filter the output, in order to control what content goes where. For example:
-injars in.jar
-outjars code_out.jar(**.class)
-outjars resources_out.jar
This configuration splits the processed output, sending **.class files to
code_out.jar, and all remaining files to resources_out.jar.
Again, the filtering can be arbitrarily complex, especially when combined with grouping input and output.
You can process several dependent or independent applications (or applets, midlets,...) in one go, in order to save time and effort. ProGuard's input and output handling offers various ways to keep the output nicely structured.
The easiest way is to specify your input jars (and/or wars, ears, zips, and directories) and a single output directory. ProGuard will then reconstruct the input in this directory, using the original jar names. For example, showing just the input and output options:
-injars application1.jar
-injars application2.jar
-injars application3.jar
-outjars processed_applications
After processing, the directory processed_applications will contain
processed versions of application jars, with their original names.
After having processed an application, e.g. ProGuard itself, you can still incrementally add other pieces of code that depend on it, e.g. the ProGuard GUI:
-injars proguardgui.jar
-outjars proguardgui_out.jar
-injars proguard.jar
-outjars proguard_out.jar
-libraryjars <java.home>/jmods/java.base.jmod(!**.jar;!module-info.class)
-applymapping proguard.map
-keep public class proguard.gui.ProGuardGUI {
public static void main(java.lang.String[]);
}
We're reading both unprocessed jars as input. Their processed contents will go
to the respective output jars. The -applymapping
option then makes sure the ProGuard part of the code gets the previously
produced obfuscation mapping. The final application will consist of the
obfuscated ProGuard jar and the additional obfuscated GUI jar.
The added code in this example is straightforward; it doesn't affect the
original code. The proguard_out.jar will be identical to the one produced in
the initial processing step. If you foresee adding more complex extensions to
your code, you should specify the options
-useuniqueclassmembernames,
-dontshrink, and
-dontoptimize in the original processing step.
These options ensure that the obfuscated base jar will always remain usable
without changes. You can then specify the base jar as a library jar:
-injars proguardgui.jar
-outjars proguardgui_out.jar
-libraryjars proguard.jar
-libraryjars <java.home>/jmods/java.base.jmod(!**.jar;!module-info.class)
-applymapping proguard.map
-keep public class proguard.gui.ProGuardGUI {
public static void main(java.lang.String[]);
}
Even if you're not interested in shrinking, optimizing, and obfuscating your midlets, as shown in the midlets example, you can still use ProGuard to preverify the class files for Java Micro Edition. ProGuard produces slightly more compact results than the traditional external preverifier.
-injars in.jar
-outjars out.jar
-libraryjars /usr/local/java/wtk2.5.2/lib/midpapi20.jar
-libraryjars /usr/local/java/wtk2.5.2/lib/cldcapi11.jar
-dontshrink
-dontoptimize
-dontobfuscate
-microedition
We're not processing the input, just making sure the class files are
preverified by targeting them at Java Micro Edition with the
-microedition option. Note that we don't need any
-keep options to specify entry points; all class files are
simply preverified.
The following options upgrade class files to Java 6, by updating their internal version numbers and preverifying them. The class files can then be loaded more efficiently by the Java 6 Virtual Machine.
-injars in.jar
-outjars out.jar
-libraryjars <java.home>/jmods/java.base.jmod(!**.jar;!module-info.class)
-dontshrink
-dontoptimize
-dontobfuscate
-target 1.6
We're not processing the input, just retargeting the class files with the
-target option. They will automatically be preverified
for Java 6 as a result. Note that we don't need any -keep options to specify
entry points; all class files are simply updated and preverified.
These options list unused classes, fields, and methods in the application
com.example.MyApplication:
-injars in.jar
-libraryjars <java.home>/jmods/java.base.jmod(!**.jar;!module-info.class)
-dontoptimize
-dontobfuscate
-dontpreverify
-printusage
-keep public class com.example.MyApplication {
public static void main(java.lang.String[]);
}
We're not specifying an output jar, just printing out some results. We're saving some processing time by skipping the other processing steps.
The java compiler inlines primitive constants and String constants (static final fields). ProGuard would therefore list such fields as not being used in
the class files that it analyzes, even if they are used in the source files.
We can add a -keepclassmembers option that
keeps those fields a priori, in order to avoid having them listed:
-keepclassmembers class * {
static final % *;
static final java.lang.String *;
}
These options print out the internal structure of all class files in the input jar:
-injars in.jar
-dontshrink
-dontoptimize
-dontobfuscate
-dontpreverify
-dump
Note how we don't need to specify the Java run-time jar, because we're not processing the input jar at all.
The traditional ProGuard configuration allows to keep a clean separation between the code and the configuration for shrinking, optimization, and obfuscation. However, it is also possible to define specific annotations, and then annotate the code to configure the processing.
You can find a set of such predefined annotations in lib/annotations.jar in
the ProGuard distribution. The corresponding ProGuard configuration (or
meta-configuration, if you prefer) is specified in
annotations/annotations.pro. With these files, you can start annotating your
code. For instance, a java source file Application.java can be annotated as
follows:
@KeepApplication
public class Application {
....
}
The ProGuard configuration file for the application can then be simplified by leveraging these annotations:
-injars in.jar
-outjars out.jar
-libraryjars <java.home>/jmods/java.base.jmod(!**.jar;!module-info.class)
-include lib/annotations.pro
The annotations are effectively replacing the application-dependent -keep
options. You may still wish to add traditional -keep
options for processing native methods,
enumerations, serializable classes, and
annotations.
The directory examples/annotations contains more examples that illustrate
some of the possibilities.