OSDN Git Service
The Android Automerger [Thu, 7 Jul 2011 13:45:57 +0000 (06:45 -0700)]
merge in ics-release history after reset to master
Elliott Hughes [Thu, 7 Jul 2011 00:26:38 +0000 (17:26 -0700)]
Turn on third-party JNI logging if the relevant bit is set.
Change-Id: I3260233516fa2d57c4876056168d31881d8666a9
Elliott Hughes [Wed, 6 Jul 2011 21:22:18 +0000 (14:22 -0700)]
Don't abort when a weak global's referent is cleared.
This also makes us less likely to output spurious warnings when
dealing with nulled-out weak globals, and lets us provide more
helpful warnings when warnings are called for.
Bug:
4991942
Change-Id: I99b88e66e07f79562da2cd9d594b93bff218d595
The Android Automerger [Mon, 4 Jul 2011 13:47:47 +0000 (06:47 -0700)]
merge in ics-release history after reset to master
jeffhao [Thu, 30 Jun 2011 23:52:28 +0000 (16:52 -0700)]
Added mutex for jdwp socket writes to prevent interleaving.
This commit includes a following change: Clean up jdwp adb/socket
and fix initialization issues.
Change-Id: I79461b92d690e42d8d1d82d31fbcacb4d5650a74
Elliott Hughes [Thu, 30 Jun 2011 20:40:52 +0000 (13:40 -0700)]
Fix native method logging to show local references rather than direct pointers.
This is necessary (but not sufficient) for debugging third-party JNI bugs.
It's the second half of the logging story, but still doesn't address the
question of "how does the developer turn on the logging?".
This removes the variant JNI bridges at the cost of adding a couple of
booleans to struct Method. Performance is about the same, except synchronized
native methods are quite a bit faster after the change.
Before:
benchmark ns linear runtime
_emptyJniMethod0 333 ==========
_emptyJniMethod6 367 ===========
_emptyJniMethod6L 921 ==============================
_emptyJniStaticMethod0 259 ========
_emptyJniStaticMethod6 287 =========
_emptyJniStaticMethod6L 873 ============================
_emptyJniStaticSynchronizedMethod0 404 =============
_emptyJniSynchronizedMethod0 452 ==============
After:
benchmark ns linear runtime
_emptyJniMethod0 344 ==========
_emptyJniMethod6 348 ==========
_emptyJniMethod6L 969 ==============================
_emptyJniStaticMethod0 265 ========
_emptyJniStaticMethod6 293 =========
_emptyJniStaticMethod6L 968 =============================
_emptyJniStaticSynchronizedMethod0 265 ========
_emptyJniSynchronizedMethod0 323 ==========
A better optimization for the case where there are reference arguments
would be to keep a list of argument indexes in the struct Method, so we
could iterate directly over those arguments that need converting to
local references. That would also let us do something about the overhead
of repeatedly looking up which local reference table and cookie to use.
But now is not the time.
Change-Id: Ie32daca1b31be057a44f1ed4b5d28d1634380e1d
Elliott Hughes [Thu, 30 Jun 2011 17:40:04 +0000 (10:40 -0700)]
Merge "Add two new JNI debugging options."
Elliott Hughes [Thu, 30 Jun 2011 17:31:31 +0000 (10:31 -0700)]
Add two new JNI debugging options.
The alwaysCheckThread option provides some backwards compatibility for apps
that misuse JNIEnv*s across threads. The logThirdPartyJni is a step towards
making it easier for third-party developers to debug their JNI errors.
Change-Id: I134374da0fe94f3fbc6b6d5aef52e3eef658aff9
The Android Automerger [Thu, 30 Jun 2011 13:44:03 +0000 (06:44 -0700)]
merge in ics-release history after reset to master
Doug Kwan [Wed, 29 Jun 2011 03:53:07 +0000 (20:53 -0700)]
Mark a variable as volatile to suppress a gcc warning.
Change-Id: Ic5263e2d7e52c72d35bf706d86bba2ad6548ce58
Elliott Hughes [Wed, 29 Jun 2011 18:17:45 +0000 (11:17 -0700)]
Merge "Make it easier to recognize JNI reference bugs."
Elliott Hughes [Wed, 29 Jun 2011 18:10:21 +0000 (11:10 -0700)]
Make it easier to recognize JNI reference bugs.
CheckJNI is still your best choice, but now, even without CheckJNI, you'll get
a strong clue:
E/dalvikvm( 4168): JNI ERROR (app bug): accessed stale local reference at index 10 (top=8)
E/dalvikvm( 4168): VM aborting
Change-Id: Ia65e1e430bc53509add19816ef06cc5dbd1b3e98
The Android Automerger [Wed, 29 Jun 2011 13:46:17 +0000 (06:46 -0700)]
merge in ics-release history after reset to master
Jeff Brown [Wed, 29 Jun 2011 03:14:46 +0000 (20:14 -0700)]
Merge "Make dual clock mode the default."
Elliott Hughes [Fri, 24 Jun 2011 22:52:55 +0000 (15:52 -0700)]
Improve JNI tracing even further.
This adds return values and general support for the primitive types.
For example:
JNI: libcore.io.Posix.writeBytes -> GetIntField((JNIEnv*)0x9ebf9a8, 0x3e00005, java.io.FileDescriptor.descriptor)
JNI: <- GetIntField returned 13
Although we still decode jclass, when a jclass is a return value we also
include the pointer. This makes it easier to follow call sequences like
this:
JNI: dalvik.system.NativeStart.main -> FindClass((JNIEnv*)0x9e3e458, "java/text/Bidi$Run")
JNI: <- FindClass returned java.lang.Class<java.text.Bidi$Run> (0x100001)
JNI: dalvik.system.NativeStart.main -> NewGlobalRef((JNIEnv*)0x9e3e458, 0x100001)
JNI: <- NewGlobalRef returned 0x100002
Here, the argument to NewGlobalRef is clearly the jclass local reference.
(Cherry-pick of
047cab15c4cb967568b25b616f02ed0f9a169430 from dalvik-dev.)
Change-Id: Ie8997b11a9cc9a0bc8f11ab3fc43f4892921316b
Elliott Hughes [Wed, 22 Jun 2011 20:24:06 +0000 (13:24 -0700)]
Improve -verbose:jni.
The old output just told you what functions were being called and made no
attempt to show you their arguments. The new output was sufficient to debug
an actual problem with an app we don't have the source to.
Still to do:
0. an easier way for third-party developers to enable this.
1. the primitive type arguments to SetIntField and so forth.
2. return values.
A few examples of the new output...
A decoded jclass:
JNI: libcore.io.Posix.readBytes called IsInstanceOf((JNIEnv*)0x9618470, 0x28100015, java.lang.Class<byte[]>)
A decoded jfieldID:
JNI: libcore.io.Posix.ioctlInt called GetIntField((JNIEnv*)0x9618470, 0x5cb00011, java.io.FileDescriptor.descriptor)
A decoded jmethodID (the FileDescriptor constructor):
JNI: libcore.io.Posix.open called NewObject((JNIEnv*)0x9780480, java.lang.Class<java.io.FileDescriptor>, java.io.FileDescriptor.<init>()V, ...)
A const char*:
JNI: libcore.io.Posix.getsockoptLinger called NewStringUTF((JNIEnv*)0x9618470, "getsockopt")
A jint release mode:
JNI: libcore.io.Posix.writeBytes called ReleaseByteArrayElements((JNIEnv*)0x9780480, 0x2700009, (void*) 0xf5f623c4, JNI_ABORT)
The -verbose:jni option now turns on a bit more output about JNI_OnLoad calls
but no longer causes any logging of calls to JNIEnv or JavaVM functions. The
old -Xjnitrace: option has been extended to enable this new tracing for the
native methods that it covers. They go very well together for debugging
purposes.
I've also made us a bit more verbose if we fail to initialize. In the longer
term I think we want to just abort if we hit any failure during startup, but
my extra logging will save us a bit of time next time we have one of these
failures (this one was caused for me by only having one half of the finalizer
watchdog change; I was missing the libcore side).
(Cherry pick of
6734b8224fb869c94e42e704ec03f2ce8483af2b from dalvik-dev.)
Change-Id: I69b7620b20620e9f06576da244520d9d83f89ab8
Jeff Brown [Tue, 28 Jun 2011 23:01:27 +0000 (16:01 -0700)]
Merge "Add a dual clock profiler tracing mode."
The Android Automerger [Tue, 28 Jun 2011 13:44:22 +0000 (06:44 -0700)]
merge in ics-release history after reset to master
jeffhao [Mon, 27 Jun 2011 22:07:17 +0000 (15:07 -0700)]
Fix debugger output of floats and doubles.
Change-Id: Idd839eef0ecb023fc4cdec20648bebbe5756d399
Jeff Brown [Sat, 25 Jun 2011 00:53:48 +0000 (17:53 -0700)]
Make dual clock mode the default.
Change-Id: Ic2f7cb8a8055319472d16be03c7ca242bf915af0
Jeff Brown [Sat, 25 Jun 2011 00:05:33 +0000 (17:05 -0700)]
Add a dual clock profiler tracing mode.
Dalvik previously supported using either the thread-cpu clock
or the real time clock as the timebase for profiler traces.
This change adds a dual clock mode where both thread-cpu time
and real time timestamps are collected.
Using dual clock mode significantly improves TraceView ability
to accurately reconstruct the global timeline of events,
particularly on SMP systems.
For now, thread-cpu mode remains the default.
Dual clock mode can be enabled by running the following command
and restarting the system server.
adb shell setprop dalvik.vm.extra-opts -Xprofile:dualclock
Change-Id: I8c0d91a99aa6829dadea328e54dc1225d9827391
The Android Automerger [Mon, 27 Jun 2011 21:24:57 +0000 (14:24 -0700)]
merge in ics-release history after reset to master
The Android Automerger [Mon, 27 Jun 2011 13:51:09 +0000 (06:51 -0700)]
merge in ics-release history after reset to master
Doug Kwan [Sun, 26 Jun 2011 19:26:18 +0000 (12:26 -0700)]
Fix a use of uninitialized variable found by gcc-4.6. I don't think this is a
real problem as we will only see an undefined value if the list has no instructions other than NOPs.
Change-Id: I055510831ca1c566e2daa2b4b2acbaa655fa735a
Doug Kwan [Sat, 25 Jun 2011 00:39:27 +0000 (17:39 -0700)]
Merge "Fix gcc-4.6 warnings."
Doug Kwan [Fri, 24 Jun 2011 22:48:32 +0000 (15:48 -0700)]
Fix gcc-4.6 warnings.
Change-Id: I62cfb8b1a39499366ee18c1904e6991567f78571
Brett Chabot [Fri, 24 Jun 2011 21:34:26 +0000 (14:34 -0700)]
Fix for adb shell hang after adb root.
Abort JDWP thread if netstate is shutting down.
Bug
4731499
Change-Id: I4e1ec9cf64d00682b1dd10e448c3665a4fc77ba8
The Android Automerger [Fri, 24 Jun 2011 20:30:33 +0000 (13:30 -0700)]
merge in ics-release history after reset to master
Elliott Hughes [Fri, 24 Jun 2011 00:29:38 +0000 (17:29 -0700)]
Throw NullPointerExceptions with detail messages.
Bug:
4905110
Change-Id: Iebcc45049c1ea14ceef6d44a19dd8cb618392101
Elliott Hughes [Thu, 23 Jun 2011 23:02:47 +0000 (16:02 -0700)]
Merge "Remove an unused internal native."
Elliott Hughes [Thu, 23 Jun 2011 22:50:13 +0000 (15:50 -0700)]
Remove an unused internal native.
Change-Id: I9a8976e7e505dfa6b641a76edc1e65be68b73339
buzbee [Thu, 23 Jun 2011 22:48:55 +0000 (15:48 -0700)]
Merge "[JIT] Fix uninitialized read (issue #
4929028)"
buzbee [Thu, 23 Jun 2011 22:20:56 +0000 (15:20 -0700)]
[JIT] Fix uninitialized read (issue #
4929028)
Found via Valgrind. This should not have caused any bad code to be
generated in this case, but good to fix. Use arrays cleared on creation.
Change-Id: I5f87c6ec74e12b9c99835695b71b6ee49505503a
Elliott Hughes [Thu, 23 Jun 2011 17:29:10 +0000 (10:29 -0700)]
Fix native methods that weren't registered via RegisterNatives.
Cherry pick of
a64af4aabf261d34eac8b5a9d92992ee70051829.
Change-Id: I43df4e33e39ccaf9e26c842f22da3391cfa17e2b
Elliott Hughes [Thu, 23 Jun 2011 18:57:43 +0000 (11:57 -0700)]
resolved conflicts for merge of
cc8c750a to master
(The transition to C++ requires an additional cast.)
Change-Id: I736f6533511d5ab2d1f05903b06d8f333adba492
Elliott Hughes [Thu, 23 Jun 2011 17:29:10 +0000 (10:29 -0700)]
Fix native methods that weren't registered via RegisterNatives.
Cherry pick of
a64af4aabf261d34eac8b5a9d92992ee70051829.
Change-Id: I43df4e33e39ccaf9e26c842f22da3391cfa17e2b
Brian Carlstrom [Wed, 22 Jun 2011 17:58:40 +0000 (10:58 -0700)]
Merge remote branch 'goog/dalvik-dev' into dalvik-dev-to-master
Elliott Hughes [Wed, 22 Jun 2011 17:30:17 +0000 (10:30 -0700)]
am
03c60a44: Merge "Undecoded jobject leaking in through attachThread"
* commit '
03c60a44861951126affa75ba8ee4233c136fe92':
Undecoded jobject leaking in through attachThread
Elliott Hughes [Wed, 22 Jun 2011 16:25:01 +0000 (09:25 -0700)]
Merge "Undecoded jobject leaking in through attachThread"
Elliott Hughes [Tue, 21 Jun 2011 21:55:09 +0000 (14:55 -0700)]
Merge "Clean up IndirectRefTable a bit." into dalvik-dev
Elliott Hughes [Tue, 21 Jun 2011 00:50:41 +0000 (17:50 -0700)]
Clean up IndirectRefTable a bit.
The main purpose here was to have slightly less unclear warnings for
JNI local reference abuse.
Change-Id: I2c6378dd0a94d8afb96a8e409f7460205e3cd315
Jesse Wilson [Tue, 21 Jun 2011 18:32:39 +0000 (11:32 -0700)]
am
bea345e1: Fix the watchdog test to handle the failure from managed code.
* commit '
bea345e1cd8acd993758366df42af7c6c9cf76d6':
Fix the watchdog test to handle the failure from managed code.
Jesse Wilson [Tue, 21 Jun 2011 17:22:19 +0000 (10:22 -0700)]
Fix the watchdog test to handle the failure from managed code.
Change-Id: I9691f414b953c7e11791069a574441b993611238
http://b/
4144865
Jesse Wilson [Tue, 21 Jun 2011 03:26:56 +0000 (20:26 -0700)]
am
0c632a84: Merge "Add isDebuggerActive API for the finalizer watchdog."
* commit '
0c632a8484839ceeffc6fdf04c310c02c7531530':
Add isDebuggerActive API for the finalizer watchdog.
Jesse Wilson [Tue, 21 Jun 2011 02:55:27 +0000 (19:55 -0700)]
Merge "Add isDebuggerActive API for the finalizer watchdog."
Jesse Wilson [Mon, 20 Jun 2011 22:09:16 +0000 (15:09 -0700)]
Add isDebuggerActive API for the finalizer watchdog.
Change-Id: Ib0afcd15327c0a3dd8c565036a8f010e318adc9e
http://b/
4144865
Elliott Hughes [Mon, 20 Jun 2011 18:53:41 +0000 (11:53 -0700)]
Don't be too specific in reference table summaries.
Implying that the examplar Class' specific type was true of all
the classes was misleading; we should just say "Class" in the
summary:
W/dalvikvm( 801): JNI local reference table summary (23 entries):
W/dalvikvm( 801): 20 of java.lang.Class (19 unique instances)
W/dalvikvm( 801): 2 of java.lang.String (2 unique instances)
W/dalvikvm( 801): 1 of java.lang.String[] (2 elements)
Change-Id: Ia961cfc689aa390c84ecbc96e44bdd64448ac051
Elliott Hughes [Mon, 20 Jun 2011 17:46:08 +0000 (10:46 -0700)]
Remove incorrect comment.
Change-Id: I3d1d84f8d36b8ee3d07e0af7e29faa72f3615e92
Elliott Hughes [Fri, 17 Jun 2011 23:57:49 +0000 (16:57 -0700)]
Fix build: rebase went wrong.
Change-Id: I390edcb3ddd52d2fb9285d2de42617dff68fdcf2
Elliott Hughes [Fri, 17 Jun 2011 22:59:16 +0000 (15:59 -0700)]
Make some of the StringObject functions member functions.
Change-Id: I72ed13c16f0cb24498772c453ba268a0f65f208a
Elliott Hughes [Fri, 17 Jun 2011 20:55:27 +0000 (13:55 -0700)]
Add and use StringAppendF.
I've also removed the dvm prefix from the existing StringPrintf.
Change-Id: I3d01f676c876957b6c6e032c674e1a06c3670021
Elliott Hughes [Fri, 17 Jun 2011 17:33:10 +0000 (10:33 -0700)]
Merge "Improve reference table dumping." into dalvik-dev
Elliott Hughes [Thu, 16 Jun 2011 23:55:24 +0000 (16:55 -0700)]
Improve reference table dumping.
1. Show the newest entry first; I always assume the top is the newest.
2. Use human-readable type names.
3. Improve the human-readable type name code to show _which_ Class (i.e.
java.lang.Class<java.lang.String> rather than just java.lang.Class).
4. Make it clear when we're reporting the number of elements in an array.
5. Show the first few characters of a string.
6. Show the length of a string if we truncate it.
(I've also removed some redundant casts and improved const-correctness.)
Example:
Last 10 entries in JNI local reference table:
16: 0x40f8ec70 java.lang.String "android.permissi... (41 chars)
15: 0x40f8d450 android.os.Parcel
14: 0x40f8eb90 java.lang.String "BlackSurface"
13: 0x408caca0 android.view.SurfaceSession
12: 0x40f8eb60 android.view.Surface
11: 0x406bc6f0 java.lang.Class<com.android.server.SystemServer>
10: 0x406c0278 java.lang.String "com/android/serv... (31 chars)
9: 0x4015d488 dalvik.system.PathClassLoader
8: 0x40148de8 java.lang.Class<java.lang.ClassLoader>
7: 0x406bc560 java.lang.String[]
JNI local reference table summary (17 entries):
6 of java.lang.Class<com.android.server.SystemServer> (5 unique instances)
5 of java.lang.String (5 unique instances)
1 of java.lang.String[]
1 of java.lang.String[] (2 elements)
1 of dalvik.system.PathClassLoader
1 of android.os.Parcel
1 of android.view.SurfaceSession
1 of android.view.Surface
Change-Id: I56494104cd0daada3ecc1e610f1c94df1e11c640
Elliott Hughes [Thu, 16 Jun 2011 21:04:10 +0000 (14:04 -0700)]
am
a7ab2182: Fix unbundled apps on pre-ICS devices.
* commit '
a7ab2182266fb9ea1e4e8d1347fc6d6da35760d4':
Fix unbundled apps on pre-ICS devices.
Elliott Hughes [Thu, 16 Jun 2011 19:56:27 +0000 (12:56 -0700)]
Fix unbundled apps on pre-ICS devices.
Change-Id: I4aa58b5395423ec2f05be3dcb1783d0e35eabb2c
Elliott Hughes [Wed, 15 Jun 2011 18:12:05 +0000 (11:12 -0700)]
Break a dependency on frameworks/base when building a host VM.
These aren't necessarily good abstractions, but they're no worse than what
we had, and having them factored out is a step in the right direction.
Change-Id: I5b839608317d2ca1ca54d8a38624fb686f2c37de
Elliott Hughes [Wed, 15 Jun 2011 16:31:44 +0000 (09:31 -0700)]
Better comments for yesterday's change.
Change-Id: I1e05fb0e209d44874101b5ca8b7c8efec6810d5f
Carl Shapiro [Wed, 15 Jun 2011 05:08:41 +0000 (22:08 -0700)]
am
cd10c009: am
f682dd64: (-s ours) am
1dd6e412: am
a028d87f: Always grow the limit address of the a heap along with its size.
* commit '
cd10c0092f7056f4d0b42bf734f8badf75d84fb9':
Always grow the limit address of the a heap along with its size.
Carl Shapiro [Wed, 15 Jun 2011 05:02:07 +0000 (22:02 -0700)]
am
f682dd64: (-s ours) am
1dd6e412: am
a028d87f: Always grow the limit address of the a heap along with its size.
* commit '
f682dd646a92f53221981a434abc06da5d249404':
Always grow the limit address of the a heap along with its size.
Carl Shapiro [Wed, 15 Jun 2011 04:59:05 +0000 (21:59 -0700)]
am
1dd6e412: am
a028d87f: Always grow the limit address of the a heap along with its size.
* commit '
1dd6e4124b5ec8b7392163cb055028678aaf042d':
Always grow the limit address of the a heap along with its size.
Carl Shapiro [Wed, 15 Jun 2011 04:56:54 +0000 (21:56 -0700)]
am
a028d87f: Always grow the limit address of the a heap along with its size.
* commit '
a028d87fb310eae8cac821183aa91567ec8cb0f9':
Always grow the limit address of the a heap along with its size.
Carl Shapiro [Wed, 15 Jun 2011 03:51:45 +0000 (20:51 -0700)]
Always grow the limit address of the a heap along with its size.
Previously, clearing the growth limit only updated the maximumSize of
the unlimited heap without updating the limit value. This caused any
operations iterated over the extent of the heap to ignore allocations
beyond the original heap size. With this change the limit is always
kept in agreement with the maximumSize.
Bug:
4539441
Change-Id: I08c8004e724549d196e1fd5d6ad22efc85b0c8d2
Carl Shapiro [Wed, 15 Jun 2011 03:31:24 +0000 (20:31 -0700)]
Normalize the include guard style.
An leading underscore followed by a capital letter is a reserved
name space in C and C++.
This change also moves any #include directives within the include
guard in some of the compiler/codegen/arm header files.
Change-Id: I9715e2c5301699d31886e61d0fe6e29483555a2a
jeffhao [Tue, 14 Jun 2011 23:37:22 +0000 (16:37 -0700)]
Dx defaults to emitting dexfiles without jumbo opcodes.
There's compatibility issues with going back and forth between versions,
and it's best to have dx default to a stable point and allow
optimizations to be controlled by switches. For jumbo opcodes to be
enabled, dx can be given the switch "--target-api=14".
Change-Id: I824092c795638ca329325b14c435bc41c1ab56d5
Elliott Hughes [Tue, 14 Jun 2011 20:51:55 +0000 (13:51 -0700)]
Allow native methods to declare they don't need a JNIEnv*.
Bug:
3069458
Change-Id: Ic9a6c562c5abf9607dd4c8a71b0d1e389e6d340b
Carl Shapiro [Tue, 14 Jun 2011 03:10:18 +0000 (20:10 -0700)]
Merge "Fix an issue regarding FindClass and threads without native methods." into dalvik-dev
Carl Shapiro [Fri, 10 Jun 2011 19:54:06 +0000 (12:54 -0700)]
Merge "Always grow the limit address of the a heap along with its size." into dalvik-dev
Carl Shapiro [Fri, 10 Jun 2011 06:45:49 +0000 (23:45 -0700)]
Do not initiate a concurrent collection if one is already running.
It is possible to cause a recursive garbage collection by writing a
program that triggers a concurrent garbage collection and initiates a
concurrent garbage collection before the garbage collection thread is
scheduled. For example
for (;;) { new byte[16 << 20]; System.gc(); }
When this condition occurs a warning is logged although such warnings
ought to be upgraded to fatal errors. With this change, when the
garbage collection thread is scheduled it first checks to see if there
is a running collection before calling down to start the collection.
Change-Id: Ia7baf5eba245bbf3fe053d3bad3f90876cad2459
Carl Shapiro [Fri, 10 Jun 2011 06:07:37 +0000 (23:07 -0700)]
Always grow the limit address of the a heap along with its size.
Previously, clearing the growth limit only updated the maximumSize of
the unlimited heap without updating the limit value. This caused any
operations iterated over the extent of the heap to ignore allocations
beyond the original heap size. With this change the limit is always
kept in agreement with the maximumSize.
Bug:
4539441
Change-Id: I08c8004e724549d196e1fd5d6ad22efc85b0c8d2
Carl Shapiro [Fri, 10 Jun 2011 05:03:57 +0000 (22:03 -0700)]
Fix an issue regarding FindClass and threads without native methods.
Chapter 4 of the JNI specification states that when FindClass is called
through the Invocation Interface and there is no associated method, the
result of ClassLoader.getBaseClassLoader() is used as the class loader.
Previously, the case where FindClass is called from a main thread was
special cased to ensure this behavior. However, threads which attach to
the VM but are not the main thread require similar treatment. With this
change those threads are similarly treated as a special case.
Change-Id: Idbe33e02a10d248262b9e9f089b033ffe05c4706
Carl Shapiro [Thu, 9 Jun 2011 07:33:49 +0000 (00:33 -0700)]
Merge "Fix a minor leaks caused by failed initializations." into dalvik-dev
Carl Shapiro [Thu, 9 Jun 2011 01:35:26 +0000 (18:35 -0700)]
Fix a minor leaks caused by failed initializations.
Change-Id: Ie0939bce7422b224f4cfde9781470f2237ede497
Elliott Hughes [Wed, 8 Jun 2011 22:54:05 +0000 (15:54 -0700)]
Tidy up some includes.
Change-Id: I5abd4f8ab06db0c77fa369fcd97d8f1259b343df
Carl Shapiro [Wed, 8 Jun 2011 20:07:44 +0000 (13:07 -0700)]
Merge "Favor Object* over void* for the heap bitmap interfaces." into dalvik-dev
Carl Shapiro [Wed, 8 Jun 2011 04:42:44 +0000 (21:42 -0700)]
Fix two issues with the card table.
Several tricks are used to adjust the base of the card table to save
instructions in the write barrier. This change corrects two
situtations where the card table did not properly account for the
effects of these tricks.
First, the card table is over allocated to ensure that the start
address can be magically aligned. When releasing the allocation these
additional bytes were not supplied to munmap creating a potential
leak.
Second, the valid card check did not use the right base address for a
containment check. The check used the allocation base which is may be
lower than the actual starting address. This would cause valid cards
at the high end of the heap to be considered invalid and invalid cards
at the low end of the heap to be considered valid.
Change-Id: I83bf4167b7c081c8626b6ff504342f42d0e776a2
Elliott Hughes [Wed, 8 Jun 2011 00:38:12 +0000 (17:38 -0700)]
Fix regular exit.
Change-Id: I653658f3b6e533a235f1108d0b625b9bdff7630c
Carl Shapiro [Wed, 8 Jun 2011 00:32:01 +0000 (17:32 -0700)]
Favor Object* over void* for the heap bitmap interfaces.
Change-Id: I615dbff3e81a1128dc3ba43d6d426c370ae3abcf
Elliott Hughes [Tue, 7 Jun 2011 23:39:20 +0000 (16:39 -0700)]
Improve "waiting on"/"waiting to lock" SIGQUIT dump info.
In particular, when we're waiting on a Class, say which class:
I(16573) - waiting on <0xf5ed54f8> (java.lang.Class<java.lang.ref.ReferenceQueue>)
versus:
I(16573) - waiting on <0xf5feda38> (a java.util.LinkedList)
Bug: http://code.google.com/p/android/issues/detail?id=17349
Change-Id: I844d02c008b1499adb02995ff3da25ba8cad0e0a
Elliott Hughes [Tue, 7 Jun 2011 22:56:08 +0000 (15:56 -0700)]
Fix gunit test build.
Change-Id: I409e75202cc6af0b0a48b81602f2ac322da7f404
Elliott Hughes [Tue, 7 Jun 2011 21:24:45 +0000 (14:24 -0700)]
Switch dvmHumanReadableDescriptor over to std::string.
(Prep work before making a change to stack dumps.)
Change-Id: I0af49b920f450fd2611e4b96e717a637483122d6
Elliott Hughes [Tue, 7 Jun 2011 17:07:11 +0000 (10:07 -0700)]
Use std::string rather than malloc/free for saneDirName.
Also use std::vector instead of the cutils/array cruft.
Change-Id: I273147335cafbac5d336955f53b0b29d015f0589
Dan Bornstein [Mon, 6 Jun 2011 18:14:07 +0000 (11:14 -0700)]
Note a possible improvement.
This section of code was left commented out because of a bug in
dx. Since the dex format version has since changed, we can now
use detection of the format version to enable the code. This
patch just notes that fact for the record.
Change-Id: I9129bc4420c17fb8c703b424a0bac9437f796210
Dan Bornstein [Thu, 2 Jun 2011 19:24:11 +0000 (12:24 -0700)]
Add some error checking and clean out some cruft.
Error checking #1: When a cached dex file can't be created, do extra
analysis to figure out (and report) why.
Error checking #2: When opening classpath entries, become sensitive
to the file extension, only trying to open files with the right
extensions and complaining explicitly if it's unrecognized.
Cruft cleaning: We've never supported finding class files in directory
hierarchies in Dalvik. Fix some related comments and clean out some
code that tried (in vain) to implement a piece of that.
Bug:
4523201
Change-Id: I05b7a8570f147955cd62229fca72b50d36703752
Brian Carlstrom [Fri, 27 May 2011 06:49:07 +0000 (23:49 -0700)]
Merge remote branch 'goog/dalvik-dev' into dalvik-dev-to-master
Dan Bornstein [Thu, 26 May 2011 19:01:03 +0000 (12:01 -0700)]
More LOG newline cleanup.
This changes all the places I could find where the log string was on the
line after its LOG call.
Change-Id: Iac6a9fcc64f46631fb093824ab60237dce1a5241
Dan Bornstein [Thu, 26 May 2011 17:46:25 +0000 (10:46 -0700)]
Prefer printf format "%#x" over "0x%x".
I exist to serve.
Change-Id: I8e2880b20eefd466da8515d5b6b0c5cb75d56169
Dan Bornstein [Thu, 26 May 2011 17:11:58 +0000 (10:11 -0700)]
Further conservation of newlines.
Friends don't let friends end LOG() strings with newlines.
Change-Id: I5a18c766c90c4ab5f03caa6acd601d34d91beb00
Dan Bornstein [Thu, 26 May 2011 05:23:29 +0000 (22:23 -0700)]
One more dex magic update.
I missed the dexdeps tool in my earlier change. This fixes it
to recognize both supported dex version numbers.
Change-Id: Ia6a26539f2ab6369ecbf3697b01b7d62a1e836b8
Dan Bornstein [Thu, 26 May 2011 05:10:37 +0000 (22:10 -0700)]
Merge "Conservation of newlines in libdex." into dalvik-dev
buzbee [Thu, 26 May 2011 01:16:51 +0000 (18:16 -0700)]
am
90f276bc: Pack interpSave struct to fix x86 & Arm diffs
* commit '
90f276bc5b8e4eeda1c4a35b2a116cbf6593d95d':
Pack interpSave struct to fix x86 & Arm diffs
buzbee [Thu, 26 May 2011 00:53:00 +0000 (17:53 -0700)]
Pack interpSave struct to fix x86 & Arm diffs
On x86, the double in JValue forces 8-byte alignment, which causes
a problem for our asm-constants. Pack it.
Change-Id: Ia53c3928a47a127fdfbb12a958111c475f83fbde
buzbee [Thu, 26 May 2011 00:40:37 +0000 (17:40 -0700)]
Merge "Pack interpSave struct to fix x86 & arm align diff" into dalvik-dev
buzbee [Thu, 26 May 2011 00:37:09 +0000 (17:37 -0700)]
Pack interpSave struct to fix x86 & arm align diff
On x86, the double in JValue forces 8-byte alignment, which causes
a problem for our asm-constants. Pack it.
Change-Id: Ie9e8bef7bb1b0ae3a502bd6fe12e831092a38812
Carl Shapiro [Thu, 26 May 2011 00:17:54 +0000 (17:17 -0700)]
Merge "Eliminate the dvmHeapSourceGetPtrFlag interface." into dalvik-dev
Dan Bornstein [Wed, 25 May 2011 23:58:39 +0000 (16:58 -0700)]
Conservation of newlines in libdex.
The log facility adds newlines automatically, and the prevailing
convention is now to omit them.
Change-Id: I14762ea4a7c24eea333f778724a0e4b6b65bae2b
Dan Bornstein [Wed, 25 May 2011 23:29:02 +0000 (16:29 -0700)]
Merge "Update dex file magic number." into dalvik-dev
Carl Shapiro [Wed, 25 May 2011 23:26:25 +0000 (16:26 -0700)]
Eliminate the dvmHeapSourceGetPtrFlag interface.
This routine served answering questions about object addresses. While
it was private to the garbage collector, its remaining use was inside
the heap profiler. Within the heap profiler, a single path was
executed to tested a pointer for containment within the zygote. In
its place, a new function, dvmIsZygoteObject, has been added to the GC
interface superseding the only use of dvmHeapSourceGetPtrFlag.
Change-Id: I347c9adb293bd6c196bbac3765015453133dc492
Dan Bornstein [Wed, 25 May 2011 20:15:47 +0000 (13:15 -0700)]
Update dex file magic number.
Even though the dex format was technically resilient with respect to
the addition of new opcodes, consensus is that the errors one sees
when trying to use a new dex file on an old build were sufficiently
inscrutable that it was worth the effort to update the version number
embedded in the dex format magic.
This change updates dx to produce the new version number when extended
opcodes are enabled (which is the default, but may be overridden by
targeting an older API level).
This also updates the vm to recognize and accept both the new current
version number as well as the immediately previous one. Note: It won't
reject an old-version file if it happens to use the new opcodes; that
would just be a gratuitous and pointless failure.
Bug:
4364986
Change-Id: If8febbb0b91c1719df4247bf69c511251362d91f
buzbee [Wed, 25 May 2011 23:00:22 +0000 (16:00 -0700)]
am
cf4a20cf: Interpreter/Debugger fix #
4479968
* commit '
cf4a20cf0cbc53f03a5b16c7152bbb29907f7108':
Interpreter/Debugger fix #
4479968
buzbee [Wed, 25 May 2011 21:21:14 +0000 (14:21 -0700)]
Interpreter/Debugger fix #
4479968
This one was tricky to track down. The underlying problem arose
with the consolidation of InterpState with Thread. Rather than
having a state structure for each instance of the interpreter, we
moved to a model that had a single thread-local struct shared by all
interpreter instances running on that thread. A portion of interpreter
state can't be shared - and thus was saved and restored on nested
invocations of the interpreter.
The bug here was that the storage for method return values was not
included in the state that needed save/retore. In normal operation,
it doesn't need to be saved - that storage isn't live across an
invoke that could trigger a nested interpreter activation. However,
when debugging, the debugger itself may hijack threads and create
new interpreter instances for its own purposed - and there is a small
window in which live retval can be trashed.
The fix is simply to move retval into the InterpSave struct.
Change-Id: Ib621824b799c5caa16fdfa8f5689a181159059df