3 <title>Android JNI Tips</title>
4 <link rel=stylesheet href="android.css">
8 <h1><a name="JNI_Tips"></a>Android JNI Tips</h1>
12 <li> <a href="#What_s_JNI_">What's JNI?</a>
14 <li> <a href="#JavaVM_and_JNIEnv">JavaVM and JNIEnv</a>
16 <li> <a href="#Threads">Threads</a>
18 <li> <a href="#jclass_jmethodID_and_jfieldID">jclass, jmethodID, and jfieldID</a>
20 <li> <a href="#local_vs_global_references">Local vs. Global References</a>
22 <li> <a href="#UTF_8_and_UTF_16_strings">UTF-8 and UTF-16 Strings</a>
24 <li> <a href="#Arrays">Primitive Arrays</a>
26 <li> <a href="#RegionCalls">Region Calls</a>
28 <li> <a href="#Exceptions">Exceptions</a>
31 <li> <a href="#Extended_checking">Extended Checking</a>
33 <li> <a href="#Native_Libraries">Native Libraries</a>
35 <li> <a href="#64bit">64-bit Considerations</a>
38 <li> <a href="#Unsupported">Unsupported Features</a>
41 <li> <a href="#FAQUnsatisfied">FAQ: UnsatisfiedLinkError</a>
43 <li> <a href="#FAQFindClass">FAQ: FindClass didn't find my class</a>
45 <li> <a href="#FAQSharing">FAQ: Sharing raw data with native code</a>
52 </p><h2><a name="What_s_JNI_"> </a> What's JNI? </h2>
55 JNI is the Java Native Interface. It defines a way for code written in the
56 Java programming language to interact with native
57 code, e.g. functions written in C/C++. It's VM-neutral, has support for loading code from
58 dynamic shared libraries, and while cumbersome at times is reasonably efficient.
60 You really should read through the
61 <a href="http://java.sun.com/javase/6/docs/technotes/guides/jni/spec/jniTOC.html">JNI spec for J2SE 1.6</a>
62 to get a sense for how JNI works and what features are available. Some
63 aspects of the interface aren't immediately obvious on
64 first reading, so you may find the next few sections handy.
65 The more detailed <i>JNI Programmer's Guide and Specification</i> can be found
66 <a href="http://java.sun.com/docs/books/jni/html/jniTOC.html">here</a>.
69 </p><h2><a name="JavaVM_and_JNIEnv"> </a> JavaVM and JNIEnv </h2>
71 JNI defines two key data structures, "JavaVM" and "JNIEnv". Both of these are essentially
72 pointers to pointers to function tables. (In the C++ version, it's a class whose sole member
73 is a pointer to a function table.) The JavaVM provides the "invocation interface" functions,
74 which allow you to create and destroy the VM. In theory you can have multiple VMs per process,
75 but Android's VM only allows one.
77 The JNIEnv provides most of the JNI functions. Your native functions all receive a JNIEnv as
81 On some VMs, the JNIEnv is used for thread-local storage. For this reason, <strong>you cannot share a JNIEnv between threads</strong>.
82 If a piece of code has no other way to get its JNIEnv, you should share
83 the JavaVM, and use JavaVM->GetEnv to discover the thread's JNIEnv.
85 The C declarations of JNIEnv and JavaVM are different from the C++
86 declarations. "jni.h" provides different typedefs
87 depending on whether it's included into ".c" or ".cpp". For this reason it's a bad idea to
88 include JNIEnv arguments in header files included by both languages. (Put another way: if your
89 header file requires "#ifdef __cplusplus", you may have to do some extra work if anything in
90 that header refers to JNIEnv.)
93 </p><h2><a name="Threads"> Threads </a></h2>
95 All VM threads are Linux threads, scheduled by the kernel. They're usually
96 started using Java language features (notably <code>Thread.start()</code>),
97 but they can also be created elsewhere and then attached to the VM. For
98 example, a thread started with <code>pthread_create</code> can be attached
99 with the JNI <code>AttachCurrentThread</code> or
100 <code>AttachCurrentThreadAsDaemon</code> functions. Until a thread is
101 attached to the VM, it has no JNIEnv, and
102 <strong>cannot make JNI calls</strong>.
104 Attaching a natively-created thread causes the VM to allocate and initialize
105 a <code>Thread</code> object, add it to the "main" <code>ThreadGroup</code>,
106 and add the thread to the set that is visible to the debugger. Calling
107 <code>AttachCurrentThread</code> on an already-attached thread is a no-op.
109 The Dalvik VM does not suspend threads executing native code. If
110 garbage collection is in progress, or the debugger has issued a suspend
111 request, the VM will pause the thread the next time it makes a JNI call.
113 Threads attached through JNI <strong>must call
114 <code>DetachCurrentThread</code> before they exit</strong>.
115 If coding this directly is awkward, in Android >= 2.0 you
116 can use <code>pthread_key_create</code> to define a destructor
117 function that will be called before the thread exits, and
118 call <code>DetachCurrentThread</code> from there. (Use that
119 key with <code>pthread_setspecific</code> to store the JNIEnv in
120 thread-local-storage; that way it'll be passed into your destructor as
124 </p><h2><a name="jclass_jmethodID_and_jfieldID"> jclass, jmethodID, and jfieldID </a></h2>
126 If you want to access an object's field from native code, you would do the following:
129 <li> Get the class object reference for the class with <code>FindClass</code>
131 <li> Get the field ID for the field with <code>GetFieldID</code>
133 <li> Get the contents of the field with something appropriate, e.g.
134 <code>GetIntField</code>
138 Similarly, to call a method, you'd first get a class object reference and then a method ID. The IDs are often just
139 pointers to internal VM data structures. Looking them up may require several string
140 comparisons, but once you have them the actual call to get the field or invoke the method
143 If performance is important, it's useful to look the values up once and cache the results
144 in your native code. Because we are limiting ourselves to one VM per process, it's reasonable
145 to store this data in a static local structure.
147 The class references, field IDs, and method IDs are guaranteed valid until the class is unloaded. Classes
148 are only unloaded if all classes associated with a ClassLoader can be garbage collected,
149 which is rare but will not be impossible in our system. Note however that
150 the <code>jclass</code>
151 is a class reference and <strong>must be protected</strong> with a call
152 to <code>NewGlobalRef</code> (see the next section).
154 If you would like to cache the IDs when a class is loaded, and automatically re-cache them
155 if the class is ever unloaded and reloaded, the correct way to initialize
156 the IDs is to add a piece of code that looks like this to the appropriate class:
160 * We use a class initializer to allow the native code to cache some
165 * A native function that looks up and caches interesting
166 * class/field/method IDs for this class. Returns false on failure.
168 native private static boolean nativeClassInit();
171 * Invoke the native initializer when the class is loaded.
174 if (!nativeClassInit())
175 throw new RuntimeException("native init failed");
179 Create a nativeClassInit method in your C/C++ code that performs the ID lookups. The code
180 will be executed once, when the class is initialized. If the class is ever unloaded and
181 then reloaded, it will be executed again. (See the implementation of java.io.FileDescriptor
182 for an example in our source tree.)
186 </p><h2><a name="local_vs_global_references"> Local vs. Global References </a></h2>
188 Every object that JNI returns is a "local reference". This means that it's valid for the
189 duration of the current native method in the current thread.
190 <strong>Even if the object itself continues to live on after the native method returns, the reference is not valid.</strong>
191 This applies to all sub-classes of <code>jobject</code>, including
192 <code>jclass</code>, <code>jstring</code>, and <code>jarray</code>.
193 (Dalvik VM will warn you about most reference mis-uses when extended JNI
197 If you want to hold on to a reference for a longer period, you must use
198 a "global" reference. The <code>NewGlobalRef</code> function takes the
199 local reference as an argument and returns a global one.
200 The global reference is guaranteed to be valid until you call
201 <code>DeleteGlobalRef</code>.
204 This pattern is commonly used when caching copies of class objects obtained
205 from <code>FindClass</code>, e.g.:
206 <p><pre>jclass* localClass = env->FindClass("MyClass");
207 jclass* globalClass = (jclass*) env->NewGlobalRef(localClass);
211 All JNI methods accept both local and global references as arguments.
212 It's possible for references to the same object to have different values;
213 for example, the return values from consecutive calls to
214 <code>NewGlobalRef</code> on the same object may be different.
215 <strong>To see if two references refer to the same object,
216 you must use the <code>IsSameObject</code> function.</strong> Never compare
217 references with "==" in native code.
219 One consequence of this is that you
220 <strong>must not assume object references are constant or unique</strong>
221 in native code. The 32-bit value representing an object may be different
222 from one invocation of a method to the next, and it's possible that two
223 different objects could have the same 32-bit value on consecutive calls. Do
224 not use <code>jobject</code> values as keys.
226 Programmers are required to "not excessively allocate" local references. In practical terms this means
227 that if you're creating large numbers of local references, perhaps while running through an array of
228 Objects, you should free them manually with
229 <code>DeleteLocalRef</code> instead of letting JNI do it for you. The
230 VM is only required to reserve slots for
231 16 local references, so if you need more than that you should either delete as you go or use
232 <code>EnsureLocalCapacity</code> to reserve more.
234 Note: method and field IDs are just 32-bit identifiers, not object
235 references, and should not be passed to <code>NewGlobalRef</code>. The raw data
236 pointers returned by functions like <code>GetStringUTFChars</code>
237 and <code>GetByteArrayElements</code> are also not objects.
239 One unusual case deserves separate mention. If you attach a native
240 thread to the VM with AttachCurrentThread, the code you are running will
241 never "return" to the VM until the thread detaches from the VM. Any local
242 references you create will have to be deleted manually unless you're going
243 to detach the thread soon.
247 </p><h2><a name="UTF_8_and_UTF_16_strings"> </a> UTF-8 and UTF-16 Strings </h2>
249 The Java programming language uses UTF-16. For convenience, JNI provides methods that work with "modified UTF-8" encoding
250 as well. (Some VMs use the modified UTF-8 internally to store strings; ours do not.) The
251 modified encoding only supports the 8- and 16-bit forms, and stores ASCII NUL values in a 16-bit encoding.
252 The nice thing about it is that you can count on having C-style zero-terminated strings,
253 suitable for use with standard libc string functions. The down side is that you cannot pass
254 arbitrary UTF-8 data into the VM and expect it to work correctly.
256 It's usually best to operate with UTF-16 strings. With our current VMs, the
257 <code>GetStringChars</code> method
258 does not require a copy, whereas <code>GetStringUTFChars</code> requires a malloc and a UTF conversion. Note that
259 <strong>UTF-16 strings are not zero-terminated</strong>, and \u0000 is allowed,
260 so you need to hang on to the string length as well as
264 <strong>Don't forget to Release the strings you Get</strong>. The
265 string functions return <code>jchar*</code> or <code>jbyte*</code>, which
266 are C-style pointers to primitive data rather than local references. They
267 are guaranteed valid until Release is called, which means they are not
268 released when the native method returns.
270 <strong>Data passed to NewStringUTF must be in "modified" UTF-8 format</strong>. A
271 common mistake is reading character data from a file or network stream
272 and handing it to <code>NewStringUTF</code> without filtering it.
273 Unless you know the data is 7-bit ASCII, you need to strip out high-ASCII
274 characters or convert them to proper "modified" UTF-8 form. If you don't,
275 the UTF-16 conversion will likely not be what you expect. The extended
276 JNI checks will scan strings and warn you about invalid data, but they
277 won't catch everything.
282 </p><h2><a name="Arrays"> </a> Primitive Arrays </h2>
284 JNI provides functions for accessing the contents of array objects.
285 While arrays of objects must be accessed one entry at a time, arrays of
286 primitives can be read and written directly as if they were declared in C.
288 To make the interface as efficient as possible without constraining
289 the VM implementation,
290 the <code>Get<PrimitiveType>ArrayElements</code> family of calls
291 allows the VM to either return a pointer to the actual elements, or
292 allocate some memory and make a copy. Either way, the raw pointer returned
293 is guaranteed to be valid until the corresponding <code>Release</code> call
294 is issued (which implies that, if the data wasn't copied, the array object
295 will be pinned down and can't be relocated as part of compacting the heap).
296 <strong>You must Release every array you Get.</strong> Also, if the Get
297 call fails, you must ensure that your code doesn't try to Release a NULL
300 You can determine whether or not the data was copied by passing in a
301 non-NULL pointer for the <code>isCopy</code> argument. This is rarely
304 The <code>Release</code> call takes a <code>mode</code> argument that can
305 have one of three values. The actions performed by the VM depend upon
306 whether it returned a pointer to the actual data or a copy of it:
310 <li>Actual: the array object is un-pinned.
311 <li>Copy: data is copied back. The buffer with the copy is freed.
313 <li><code>JNI_COMMIT</code>
315 <li>Actual: does nothing.
316 <li>Copy: data is copied back. The buffer with the copy
317 <strong>is not freed</strong>.
319 <li><code>JNI_ABORT</code>
321 <li>Actual: the array object is un-pinned. Earlier
322 writes are <strong>not</strong> aborted.
323 <li>Copy: the buffer with the copy is freed; any changes to it are lost.
327 One reason for checking the <code>isCopy</code> flag is to know if
328 you need to call <code>Release</code> with <code>JNI_COMMIT</code>
329 after making changes to an array — if you're alternating between making
330 changes and executing code that uses the contents of the array, you may be
332 skip the no-op commit. Another possible reason for checking the flag is for
333 efficient handling of <code>JNI_ABORT</code>. For example, you might want
334 to get an array, modify it in place, pass pieces to other functions, and
335 then discard the changes. If you know that JNI is making a new copy for
336 you, there's no need to create another "editable" copy. If JNI is passing
337 you the original, then you do need to make your own copy.
339 Some have asserted that you can skip the <code>Release</code> call if
340 <code>*isCopy</code> is false. This is not the case. If no copy buffer was
341 allocated, then the original memory must be pinned down and can't be moved by
342 the garbage collector.
344 Also note that the <code>JNI_COMMIT</code> flag does NOT release the array,
345 and you will need to call <code>Release</code> again with a different flag
351 </p><h2><a name="RegionCalls"> Region Calls </a></h2>
354 There is an alternative to calls like <code>Get<Type>ArrayElements</code>
355 and <code>GetStringChars</code> that may be very helpful when all you want
356 to do is copy data in or out. Consider the following:
358 jbyte* data = env->GetByteArrayElements(array, NULL);
360 memcpy(buffer, data, len);
361 env->ReleaseByteArrayElements(array, data, JNI_ABORT);
365 This grabs the array, copies the first <code>len</code> byte
366 elements out of it, and then releases the array. Depending upon the VM
367 policies the <code>Get</code> call will either pin or copy the array contents.
368 We copy the data (for perhaps a second time), then call Release; in this case
369 we use <code>JNI_ABORT</code> so there's no chance of a third copy.
371 We can accomplish the same thing with this:
373 env->GetByteArrayRegion(array, 0, len, buffer);
376 This has several advantages:
378 <li>Requires one JNI call instead of 2, reducing overhead.
379 <li>Doesn't require pinning or extra data copies.
380 <li>Reduces the risk of programmer error — no risk of forgetting
381 to call <code>Release</code> after something fails.
384 Similarly, you can use the <code>Set<Type>ArrayRegion</code> call
385 to copy data into an array, and <code>GetStringRegion</code> or
386 <code>GetStringUTFRegion</code> to copy characters out of a
390 </p><h2><a name="Exceptions"> Exceptions </a></h2>
392 <strong>You may not call most JNI functions while an exception is pending.</strong>
393 Your code is expected to notice the exception (via the function's return value,
394 <code>ExceptionCheck()</code>, or <code>ExceptionOccurred()</code>) and return,
395 or clear the exception and handle it.
397 The only JNI functions that you are allowed to call while an exception is
402 <li>DeleteWeakGlobalRef
405 <li>ExceptionDescribe
406 <li>ExceptionOccurred
410 <li>Release<PrimitiveType>ArrayElements
411 <li>ReleasePrimitiveArrayCritical
412 <li>ReleaseStringChars
413 <li>ReleaseStringCritical
414 <li>ReleaseStringUTFChars
417 Many JNI calls can throw an exception, but often provide a simpler way
418 of checking for failure. For example, if <code>NewString</code> returns
419 a non-NULL value, you don't need to check for an exception. However, if
420 you call a method (using a function like <code>CallObjectMethod</code>),
421 you must always check for an exception, because the return value is not
422 going to be valid if an exception was thrown.
424 Note that exceptions thrown by interpreted code do not "leap over" native code,
425 and C++ exceptions thrown by native code are not handled by Dalvik.
426 The JNI <code>Throw</code> and <code>ThrowNew</code> instructions just
427 set an exception pointer in the current thread. Upon returning to the VM from
428 native code, the exception will be noted and handled appropriately.
430 Native code can "catch" an exception by calling <code>ExceptionCheck</code> or
431 <code>ExceptionOccurred</code>, and clear it with
432 <code>ExceptionClear</code>. As usual,
433 discarding exceptions without handling them can lead to problems.
435 There are no built-in functions for manipulating the Throwable object
436 itself, so if you want to (say) get the exception string you will need to
437 find the Throwable class, look up the method ID for
438 <code>getMessage "()Ljava/lang/String;"</code>, invoke it, and if the result
439 is non-NULL use <code>GetStringUTFChars</code> to get something you can
440 hand to printf or a LOG macro.
444 </p><h2><a name="Extended_checking"> Extended Checking </a></h2>
446 JNI does very little error checking. Calling <code>SetIntField</code>
447 on an Object field will succeed, even if the field is marked
448 <code>private</code> and <code>final</code>. The
449 goal is to minimize the overhead on the assumption that, if you've written it in native code,
450 you probably did it for performance reasons.
452 In Dalvik, you can enable additional checks by setting the
453 "<code>-Xcheck:jni</code>" flag. If the flag is set, the VM directs
454 the JavaVM and JNIEnv pointers to a different table of functions.
455 These functions perform an extended series of checks before calling the
456 standard implementation.
459 The additional tests include:
463 <li> Check for null pointers where not allowed.
465 <li> Verify argument type correctness (jclass is a class object,
466 jfieldID points to field data, jstring is a java.lang.String).
468 <li> Field type correctness, e.g. don't store a HashMap in a String field.
470 <li> Ensure jmethodID is appropriate when making a static or virtual
473 <li> Check to see if an exception is pending on calls where pending exceptions are not legal.
475 <li> Check for calls to inappropriate functions between Critical get/release calls.
477 <li> Check that JNIEnv structs aren't being shared between threads.
480 <li> Make sure local references aren't used outside their allowed lifespan.
482 <li> UTF-8 strings contain only valid "modified UTF-8" data.
485 <p>Accessibility of methods and fields (i.e. public vs. private) is not
488 For a description of how to enable CheckJNI for Android apps, see
489 <a href="embedded-vm-control.html">Controlling the Embedded VM</a>.
490 It's currently enabled by default in the Android emulator and on
491 "engineering" device builds.
494 JNI checks can be modified with the <code>-Xjniopts</code> command-line
495 flag. Currently supported values include:
499 <dd>When set, any function that can return a copy of the original data
500 (array of primitive values, UTF-16 chars) will always do so. The buffers
501 are over-allocated and surrounded with a guard pattern to help identify
502 code writing outside the buffer, and the contents are erased before the
503 storage is freed to trip up code that uses the data after calling Release.
504 This will have a noticeable performance impact on some applications.
506 <dd>By default, JNI "warnings" cause the VM to abort. With this flag
512 </p><h2><a name="Native_Libraries"> Native Libraries </a></h2>
514 You can load native code from shared libraries with the standard
515 <code>System.loadLibrary()</code> call. The
516 preferred way to get at your native code is:
519 <li> Call <code>System.loadLibrary()</code> from a static class
520 initializer. (See the earlier example, where one is used to call
521 <code>nativeClassInit()</code>.) The argument is the "undecorated"
522 library name, e.g. to load "libfubar.so" you would pass in "fubar".
525 <li> Provide a native function: <code><strong>jint JNI_OnLoad(JavaVM* vm, void* reserved)</strong></code>
527 <li>In <code>JNI_OnLoad</code>, register all of your native methods. You
529 the methods "static" so the names don't take up space in the symbol table
534 The <code>JNI_OnLoad</code> function should look something like this if
536 </p><blockquote><pre>jint JNI_OnLoad(JavaVM* vm, void* reserved)
539 if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_6) != JNI_OK)
542 /* get class with (*env)->FindClass */
543 /* register methods with (*env)->RegisterNatives */
545 return JNI_VERSION_1_6;
549 You can also call <code>System.load()</code> with the full path name of the
550 shared library. For Android apps, you may find it useful to get the full
551 path to the application's private data storage area from the context object.
553 This is the recommended approach, but not the only approach. The VM does
554 not require explicit registration, nor that you provide a
555 <code>JNI_OnLoad</code> function.
556 You can instead use "discovery" of native methods that are named in a
557 specific way (see <a href="http://java.sun.com/javase/6/docs/technotes/guides/jni/spec/design.html#wp615">
558 the JNI spec</a> for details), though this is less desirable.
559 It requires more space in the shared object symbol table,
560 loading is slower because it requires string searches through all of the
561 loaded shared libraries, and if a method signature is wrong you won't know
562 about it until the first time the method is actually used.
564 One other note about <code>JNI_OnLoad</code>: any <code>FindClass</code>
565 calls you make from there will happen in the context of the class loader
566 that was used to load the shared library. Normally <code>FindClass</code>
567 uses the loader associated with the method at the top of the interpreted
568 stack, or if there isn't one (because the thread was just attached to
569 the VM) it uses the "system" class loader. This makes
570 <code>JNI_OnLoad</code> a convenient place to look up and cache class
575 </p><h2><a name="64bit"> 64-bit Considerations </a></h2>
578 Android is currently expected to run on 32-bit platforms. In theory it
579 could be built for a 64-bit system, but that is not a goal at this time.
580 For the most part this isn't something that you will need to worry about
581 when interacting with native code,
582 but it becomes significant if you plan to store pointers to native
583 structures in integer fields in an object. To support architectures
584 that use 64-bit pointers, <strong>you need to stash your native pointers in a
585 <code>long</code> field rather than an <code>int</code></strong>.
588 </p><h2><a name="Unsupported"> Unsupported Features </a></h2>
589 <p>All JNI 1.6 features are supported, with the following exceptions:
591 <li><code>DefineClass</code> is not implemented. Dalvik does not use
592 Java bytecodes or class files, so passing in binary class data
593 doesn't work. Translation facilities may be added in a future
594 version of the VM.</li>
595 <li>"Weak global" references are implemented, but may only be passed
596 to <code>NewLocalRef</code>, <code>NewGlobalRef</code>, and
597 <code>DeleteWeakGlobalRef</code>. (The spec strongly encourages
598 programmers to create hard references to weak globals before doing
599 anything with them, so this should not be at all limiting.)</li>
600 <li><code>GetObjectRefType</code> (new in 1.6) is implemented but not fully
601 functional — it can't always tell the difference between "local" and
602 "global" references.</li>
605 <p>For backward compatibility, you may need to be aware of:
607 <li>Until 2.0 ("Eclair"), the '$' character was not properly
608 converted to "_00024" during searches for method names. Working
609 around this requires using explicit registration or moving the
610 native methods out of inner classes.
611 <li>Until 2.0, it was not possible to use a <code>pthread_key_create</code>
612 destructor function to avoid the VM's "thread must be detached before
613 exit" check. (The VM also uses a pthread key destructor function,
614 so it'd be a race to see which gets called first.)
615 <li>"Weak global" references were not implemented until 2.2 ("Froyo").
616 Older VMs will vigorously reject attempts to use them. You can use
617 the Android platform version constants to test for support.
621 </p><h2><a name="FAQUnsatisfied"> FAQ: UnsatisfiedLinkError </a></h2>
623 When working on native code it's not uncommon to see a failure like this:
624 <pre>java.lang.UnsatisfiedLinkError: Library foo not found</pre>
626 In some cases it means what it says — the library wasn't found. In
627 other cases the library exists but couldn't be opened by dlopen(), and
628 the details of the failure can be found in the exception's detail message.
630 Common reasons why you might encounter "library not found" exceptions:
632 <li>The library doesn't exist or isn't accessible to the app. Use
633 <code>adb shell ls -l <path></code> to check its presence
635 <li>The library wasn't built with the NDK. This can result in
636 dependencies on functions or libraries that don't exist on the device.
639 Another class of <code>UnsatisfiedLinkError</code> failures looks like:
640 <pre>java.lang.UnsatisfiedLinkError: myfunc
641 at Foo.myfunc(Native Method)
642 at Foo.main(Foo.java:10)</pre>
644 In logcat, you'll see:
645 <pre>W/dalvikvm( 880): No implementation found for native LFoo;.myfunc ()V</pre>
647 This means that the VM tried to find a matching method but was unsuccessful.
648 Some common reasons for this are:
650 <li>The library isn't getting loaded. Check the logcat output for
651 messages about library loading.
652 <li>The method isn't being found due to a name or signature mismatch. This
653 is commonly caused by:
655 <li>For lazy method lookup, failing to declare C++ functions
656 with <code>extern C</code>. You can use <code>arm-eabi-nm</code>
657 to see the symbols as they appear in the library; if they look
658 mangled (e.g. <code>_Z15Java_Foo_myfuncP7_JNIEnvP7_jclass</code>
659 rather than <code>Java_Foo_myfunc</code>) then you need to
660 adjust the declaration.
661 <li>For explicit registration, minor errors when entering the
662 method signature. Make sure that what you're passing to the
663 registration call matches the signature in the log file.
664 Remember that 'B' is <code>byte</code> and 'Z' is <code>boolean</code>.
665 Class name components in signatures start with 'L', end with ';',
666 use '/' to separate package/class names, and use '$' to separate
668 (e.g. <code>Ljava/util/Map$Entry;</code>).
672 Using <code>javah</code> to automatically generate JNI headers may help
676 </p><h2><a name="FAQFindClass"> FAQ: FindClass didn't find my class </a></h2>
678 Make sure that the class name string has the correct format. JNI class
679 names start with the package name and are separated with slashes,
680 e.g. <code>java/lang/String</code>. If you're looking up an array class,
681 you need to start with the appropriate number of square brackets and
682 must also wrap the class with 'L' and ';', so a one-dimensional array of
683 <code>String</code> would be <code>[Ljava/lang/String;</code>.
685 If the class name looks right, you could be running into a class loader
686 issue. <code>FindClass</code> wants to start the class search in the
687 class loader associated with your code. It examines the VM call stack,
688 which will look something like:
689 <pre> Foo.myfunc(Native Method)
690 Foo.main(Foo.java:10)
691 dalvik.system.NativeStart.main(Native Method)</pre>
693 The topmost method is <code>Foo.myfunc</code>. <code>FindClass</code>
694 finds the <code>ClassLoader</code> object associated with the <code>Foo</code>
697 This usually does what you want. You can get into trouble if you
698 create a thread outside the VM (perhaps by calling <code>pthread_create</code>
699 and then attaching it to the VM with <code>AttachCurrentThread</code>).
700 Now the stack trace looks like this:
701 <pre> dalvik.system.NativeStart.run(Native Method)</pre>
703 The topmost method is <code>NativeStart.run</code>, which isn't part of
704 your application. If you call <code>FindClass</code> from this thread, the
705 VM will start in the "system" class loader instead of the one associated
706 with your application, so attempts to find app-specific classes will fail.
708 There are a few ways to work around this:
710 <li>Do your <code>FindClass</code> lookups once, in
711 <code>JNI_OnLoad</code>, and cache the class references for later
712 use. Any <code>FindClass</code> calls made as part of executing
713 <code>JNI_OnLoad</code> will use the class loader associated with
714 the function that called <code>System.loadLibrary</code> (this is a
715 special rule, provided to make library initialization more convenient).
716 If your app code is loading the library, <code>FindClass</code>
717 will use the correct class loader.
718 <li>Pass an instance of the class into the functions that need
719 it, e.g. declare your native method to take a Class argument and
720 then pass <code>Foo.class</code> in.
721 <li>Cache a reference to the <code>ClassLoader</code> object somewhere
722 handy, and issue <code>loadClass</code> calls directly. This requires
729 </p><h2><a name="FAQSharing"> FAQ: Sharing raw data with native code </a></h2>
731 You may find yourself in a situation where you need to access a large
732 buffer of raw data from code written in Java and C/C++. Common examples
733 include manipulation of bitmaps or sound samples. There are two
736 You can store the data in a <code>byte[]</code>. This allows very fast
737 access from code written in Java. On the native side, however, you're
738 not guaranteed to be able to access the data without having to copy it. In
739 some implementations, <code>GetByteArrayElements</code> and
740 <code>GetPrimitiveArrayCritical</code> will return actual pointers to the
741 raw data in the managed heap, but in others it will allocate a buffer
742 on the native heap and copy the data over.
744 The alternative is to store the data in a direct byte buffer. These
745 can be created with <code>java.nio.ByteBuffer.allocateDirect</code>, or
746 the JNI <code>NewDirectByteBuffer</code> function. Unlike regular
747 byte buffers, the storage is not allocated on the managed heap, and can
748 always be accessed directly from native code (get the address
749 with <code>GetDirectBufferAddress</code>). Depending on how direct
750 byte buffer access is implemented in the VM, accessing the data from code
751 written in Java can be very slow.
753 The choice of which to use depends on two factors:
755 <li>Will most of the data accesses happen from code written in Java
757 <li>If the data is eventually being passed to a system API, what form
758 must it be in? (For example, if the data is eventually passed to a
759 function that takes a byte[], doing processing in a direct
760 <code>ByteBuffer</code> might be unwise.)
762 If there's no clear winner, use a direct byte buffer. Support for them
763 is built directly into JNI, and access to them from code written in
764 Java can be made faster with VM improvements.
767 <address>Copyright © 2008 The Android Open Source Project</address>