OSDN Git Service
Mathieu Chartier [Thu, 11 Jun 2015 21:45:53 +0000 (21:45 +0000)]
Merge "Move image intern table into image" into mnc-dev
Mathieu Chartier [Wed, 10 Jun 2015 00:50:29 +0000 (17:50 -0700)]
Move image intern table into image
Previously we recreated this intern table during runtime startup.
This added 50-100ms of boot time.
Fixed bug where we didn't copy over hashcodes into the image.
Deleted some stale code.
Bug:
20727525
Bug:
19569780
Change-Id: I08959e9aa2a73cedb52f393033e2ffea3a26e76b
Hiroshi Yamauchi [Thu, 11 Jun 2015 00:46:10 +0000 (00:46 +0000)]
Merge "Reset GC performance stats at zygote fork." into mnc-dev
Lazar Trsic [Wed, 10 Jun 2015 14:05:46 +0000 (16:05 +0200)]
[MIPS] Fix method tracing for mips64
Reg $v0 is being clobbered on using
SETUP_REFS_ONLY_CALLEE_SAVE_FRAME inside instrumentation_exit()
trampoline, when falling through from instrumentation_entry().
To fix, use $t1 for ArtMethod* fetching inside SETUP_ macros.
Also fix some unrelated comment errors.
Bug:
21555893
(cherry picked from commit
84bc06e30ba12c3ff07e577c52b63b9df162af7e)
Change-Id: I4ab169f75988c4f0d924416a30196f21c1a043a3
Hiroshi Yamauchi [Thu, 11 Jun 2015 00:20:54 +0000 (17:20 -0700)]
Reset GC performance stats at zygote fork.
So GCs before a zygote fork won't be attributed to an app.
Bug:
21491908
Change-Id: Ib37bc587e0f039ef8faeabe63dec19de49501863
Vladimir Marko [Mon, 8 Jun 2015 15:39:02 +0000 (16:39 +0100)]
Quick: Fix LoopRepeatingTopologicalSortIterator.
Always push the loop head on the loop head stack. This fixes
a bug where we failed to return to an unnatural loop head to
recalculate its GVN data.
Bug:
17410955
(cherry picked from commit
67c8c942e9dfcabd548351db75e6d3b8b5165afa)
Change-Id: I44b9a17cbcd7307d1cc70ac564b99e29803723c5
Mathieu Chartier [Mon, 8 Jun 2015 18:33:25 +0000 (18:33 +0000)]
Merge "More LOW_4G optimizations" into mnc-dev
Andreas Gampe [Sun, 7 Jun 2015 21:35:46 +0000 (21:35 +0000)]
Merge "Revert "Revert "Turn off duplicate-classes checking""" into mnc-dev
Andreas Gampe [Sat, 6 Jun 2015 03:53:37 +0000 (03:53 +0000)]
Merge "ART: Add SetSpecialSignalHandlerFn to version-script.txt" into mnc-dev
Andreas Gampe [Sat, 6 Jun 2015 03:45:31 +0000 (20:45 -0700)]
ART: Add SetSpecialSignalHandlerFn to version-script.txt
Expose the new function from the sigchainlib implementation.
Bug:
20217701
(cherry picked from commit
e089920eb0abc10eca708dbf5993134aa7e8d9b0)
Change-Id: I7d21025595775e057ea703236ca88ceb07173136
Vladimir Marko [Sat, 6 Jun 2015 00:06:18 +0000 (00:06 +0000)]
am
b8154d5c: resolved conflicts for merge of
e8a4b652 to lmp-mr1-ub-dev
* commit '
b8154d5cc707921fadf92fedc65a66a3e8901126':
Revert "Quick: Create GC map based on compiler data. DO NOT MERGE"
Vladimir Marko [Sat, 6 Jun 2015 00:06:16 +0000 (00:06 +0000)]
am
5d3acebd: resolved conflicts for merge of
c8d000a1 to lmp-mr1-ub-dev
* commit '
5d3acebd1fe859cbf6fb3831815eb7f3c1f3e442':
Revert "Quick: Fix "select" pattern to update data used for GC maps. DO NOT MERGE"
Vladimir Marko [Sat, 6 Jun 2015 00:06:13 +0000 (00:06 +0000)]
am
e279c0c4: resolved conflicts for merge of
fad2cbf9 to lmp-mr1-ub-dev
* commit '
e279c0c4e9ecc464ab10c9d9cd3404c3b0ed6845':
Quick: Fix "select" pattern to update data used for GC maps. DO NOT MERGE
Vladimir Marko [Sat, 6 Jun 2015 00:06:12 +0000 (00:06 +0000)]
am
fb695b34: resolved conflicts for merge of
7cc8f9aa to lmp-mr1-ub-dev
* commit '
fb695b34268bb7f9270f07ecaac554e49b5b6a92':
Quick: Create GC map based on compiler data. DO NOT MERGE
Mathieu Chartier [Fri, 5 Jun 2015 20:21:05 +0000 (13:21 -0700)]
Use runFinalizationWithTimeout for native allocations
Prevents deadlocks by not waiting longer than 250ms for finalizers
to complete.
Bug:
21544853
Change-Id: I57b2f7ae8b74185922eb3c15ba0ab71a4d2348aa
Mathieu Chartier [Thu, 4 Jun 2015 23:22:41 +0000 (16:22 -0700)]
More LOW_4G optimizations
Eagerly try to map at an address when it doesn't overlap with
the ART known maps.
Reduces time spent creating the heap on N9 from ~100ms to ~10ms.
Bug:
20727525
Change-Id: Ie8982ca9d1f3158104b2b53b992c3a28841fc443
Richard Uhler [Tue, 7 Apr 2015 17:56:50 +0000 (10:56 -0700)]
Add atrace calls to runtime start and heap create
Bug:
20727525
Change-Id: I91a74b793fb9eda8ed580244a6a5fd313ef2eb27
Mathieu Chartier [Thu, 4 Jun 2015 20:42:14 +0000 (13:42 -0700)]
Add optimization for LOW_4G allocator
The optimization uses the maps_ field to skip over regions which we
know wont be large enough for the request.
Reduces time to create heap from 500-700ms to 100ms on N9.
Bug:
20727525
Change-Id: I4fa92d58c2e135ba31a03ababff46669089bb542
Vladimir Marko [Thu, 4 Jun 2015 10:14:23 +0000 (11:14 +0100)]
resolved conflicts for merge of
e8a4b652 to lmp-mr1-ub-dev
Vladimir Marko [Thu, 4 Jun 2015 10:04:05 +0000 (11:04 +0100)]
resolved conflicts for merge of
c8d000a1 to lmp-mr1-ub-dev
Vladimir Marko [Thu, 4 Jun 2015 09:59:53 +0000 (10:59 +0100)]
resolved conflicts for merge of
fad2cbf9 to lmp-mr1-ub-dev
Vladimir Marko [Thu, 4 Jun 2015 09:53:59 +0000 (10:53 +0100)]
resolved conflicts for merge of
7cc8f9aa to lmp-mr1-ub-dev
Change-Id: I6203b6d8be029440f635b9575559ba6add9a46d6
Vladimir Marko [Thu, 4 Jun 2015 09:48:26 +0000 (09:48 +0000)]
Merge "Revert "Quick: Create GC map based on compiler data. DO NOT MERGE"" into cw-d-mr1-dev
Vladimir Marko [Thu, 4 Jun 2015 09:47:47 +0000 (09:47 +0000)]
Revert "Quick: Create GC map based on compiler data. DO NOT MERGE"
This reverts commit
7cc8f9aa1349fd6cb0814a653ee2d1164a7fb9f7.
Change-Id: Iadb4462bf8e834c6a847c01ee6eb332a325de22c
Vladimir Marko [Thu, 4 Jun 2015 09:47:21 +0000 (09:47 +0000)]
Revert "Quick: Fix "select" pattern to update data used for GC maps. DO NOT MERGE"
This reverts commit
fad2cbf97c71b9742ccd88cc1a5ba13fa918e677.
Change-Id: I175dd9e49014b71a300d987678032bd624a99cf1
Mathieu Chartier [Thu, 4 Jun 2015 00:32:42 +0000 (17:32 -0700)]
Fix valgrind large_object_space_test
Also some cleanup.
Change-Id: I9c1a8093e6356f1b52e332009429b77fa5c1d448
Andreas Gampe [Wed, 3 Jun 2015 22:43:16 +0000 (15:43 -0700)]
Revert "Revert "Turn off duplicate-classes checking""
This reverts commit
18a1827a159f1b235f3fcc934f428059185f550e.
Bug:
21333911
Change-Id: I8cb96efbf7151b238f3ee137a938b20ddfaf0ebc
Mathieu Chartier [Wed, 3 Jun 2015 20:32:15 +0000 (13:32 -0700)]
Add a way to determine if a large object is a zygote object
Also fix a slight memory leak in LargeObjectMapSpace.
Bug:
20674158
Change-Id: I2416df484e5b84a8c5cc0b5664c8cb102dc235f6
Andreas Gampe [Wed, 3 Jun 2015 20:27:44 +0000 (20:27 +0000)]
Merge "ART: Refactor Thread::Init" into mnc-dev
Andreas Gampe [Tue, 2 Jun 2015 05:29:51 +0000 (22:29 -0700)]
ART: Refactor Thread::Init
This refactor allows the parent thread to allocate the JNIEnvExt
for the child (with a fallback in place in Init). This allows to
throw an OOME in CreateNativeThread instead of aborting in the
child.
Bug:
21291279
Change-Id: Iccc1a5c202999f5bfacec706d9833e53135ba2fa
Vladimir Marko [Wed, 25 Mar 2015 11:13:39 +0000 (11:13 +0000)]
Quick: Fix "select" pattern to update data used for GC maps. DO NOT MERGE
Follow-up to
https://android-review.googlesource.com/143222
(cherry picked from commit
6e07183e822a32856da9eb60006989496e06a9cc)
Change-Id: I916743c845d9568063cd6a4b2ef71e9cbc43dee8
Vladimir Marko [Fri, 20 Mar 2015 12:47:30 +0000 (12:47 +0000)]
Quick: Create GC map based on compiler data. DO NOT MERGE
The Quick compiler and verifier sometimes disagree on dalvik
register types (fp/core/ref) for 0/null constants and merged
registers involving 0/null constants. Since the verifier is
more lenient it can mark a register as a reference for GC
where Quick considers it a floating point register or a dead
register (which would have a ref/fp conflict if not dead).
If the compiler used an fp register to hold the zero value,
the core register or stack location used by GC based on the
verifier data can hold an invalid value.
Previously, as a workaround we stored the fp zero value also
in the stack location or core register where GC would look
for it. This wasn't precise and may have missed some cases.
To fix this properly, we now generate GC maps based on the
compiler's notion of references if register promotion is
enabled.
Bug: https://code.google.com/p/android/issues/detail?id=147187
(cherry picked from commit
767c752fddc64e280dba507457e4f06002b5f678)
Change-Id: Id75428fd0a2f6bdd2ccb20ce75cdeab01150e455
Mathieu Chartier [Wed, 3 Jun 2015 18:08:22 +0000 (18:08 +0000)]
Merge "Emit instance fields for java.lang.Object in hprof" into mnc-dev
Mathieu Chartier [Wed, 3 Jun 2015 17:51:13 +0000 (10:51 -0700)]
Emit instance fields for java.lang.Object in hprof
Previously we omitted these fields.
Bug:
20726205
Change-Id: I81c55919a4bd015bc0a4e7259a41a05dab534e11
Andreas Gampe [Wed, 3 Jun 2015 04:01:45 +0000 (21:01 -0700)]
ART: Prune FindArrayClass cache in image writer
The ClassLinker cache speeds up FindArrayClass requests, but all
entries are roots. It is possible that an entry is a non-image
class when creating the boot image, artificially keeping the
class around.
Bug:
21596650
Change-Id: Ief9b439945d0e293a3cb5dcddfeb189b5e174f06
Mathieu Chartier [Wed, 3 Jun 2015 00:16:33 +0000 (00:16 +0000)]
Merge "Add a way to SuspendAll for a long duration" into mnc-dev
Mathieu Chartier [Tue, 2 Jun 2015 23:38:29 +0000 (16:38 -0700)]
Fix tracing
Move back to 32 bit method IDs, add a bijective map for method IDs.
Also some cleanup.
Bug:
19264997
Change-Id: Icdd36591df53ff975d30b9000cfe67d3ae8c51d5
Mathieu Chartier [Tue, 2 Jun 2015 20:40:12 +0000 (13:40 -0700)]
Add more info for attempting to delete non-JNI local reference
Now we print the stack trace of the method which tried to remove the
non-JNI local reference.
Added test, example output:
Attempt to remove non-JNI local reference, dumping thread
"main" prio=5 tid=1 Runnable
| group="main" sCount=0 dsCount=0 obj=0x12c2b400 self=0xf878af28
| sysTid=32662 nice=0 cgrp=default sched=0/0 handle=0xf73df700
| state=R schedstat=(
154297228 636505 121 ) utm=5 stm=10 core=11 HZ=100
| stack=0xff6f2000-0xff6f4000 stackSize=7MB
| held mutexes= "mutator lock"(shared held)
native: #00 pc
00455a7f /out/host/linux-x86/lib/libartd.so (art::DumpNativeStack(std::__1::basic_ostream<char, std::__1::char_traits<char> >&, int, char const*, art::ArtMethod*, void*)+191)
native: #01 pc
00425d79 /out/host/linux-x86/lib/libartd.so (art::Thread::DumpStack(std::__1::basic_ostream<char, std::__1::char_traits<char> >&) const+473)
native: #02 pc
0028c555 /out/host/linux-x86/lib/libartd.so (art::IndirectReferenceTable::Remove(unsigned int, void*)+421)
native: #03 pc
003319c8 /out/host/linux-x86/lib/libartd.so (art::JNI::DeleteLocalRef(_JNIEnv*, _jobject*)+56)
native: #04 pc
0012db14 /out/host/linux-x86/lib/libartd.so (art::CheckJNI::DeleteRef(char const*, _JNIEnv*, _jobject*, art::IndirectRefKind)+292)
native: #05 pc
00122176 /out/host/linux-x86/lib/libartd.so (art::CheckJNI::DeleteLocalRef(_JNIEnv*, _jobject*)+54)
native: #06 pc
00005724 /out/host/linux-x86/lib/libarttest.so (Java_Main_removeLocalObject+36)
native: #07 pc
00004d6e /tmp/test-32506/dalvik-cache/x86/tmp@mathieuc@test-32506@004-JniTest.jar@classes.dex (void Main.removeLocalObject(java.lang.Object)+130)
at Main.removeLocalObject(Native method)
at Main.testRemoveLocalObject(Main.java:118)
at Main.main(Main.java:37)
Bug:
20683465
Change-Id: I56b99c77ae7020cabf1aaeb1d5317b1d5ba940e5
Mathieu Chartier [Tue, 2 Jun 2015 18:42:18 +0000 (11:42 -0700)]
Add a way to SuspendAll for a long duration
Hprof uses this mode to not cause thread suspend timeouts if GC tries
while the hprof dump is running.
Bug:
21063989
Change-Id: Ic6304620afd1489719a7e0e4299f829c90fe27cc
Mathieu Chartier [Tue, 2 Jun 2015 02:12:36 +0000 (19:12 -0700)]
Add clear cache for JIT
(cherry picked from commit
5783a74163878095475596fabc0b9cb04bec467b)
Bug:
21572270
Change-Id: I711a9ee2d9688a8e4b59380c401f10f42c6c068e
Mathieu Chartier [Tue, 2 Jun 2015 16:56:04 +0000 (09:56 -0700)]
Update IMT for stale miranda methods
Fixes occasional crash in dex2oat.
Bug:
19264997
Change-Id: Icb1a323d8c44fa40309486e17aec56d2c2588e67
Mathieu Chartier [Tue, 2 Jun 2015 16:44:43 +0000 (16:44 +0000)]
Merge "Move mirror::ArtMethod to native" into mnc-dev
Mathieu Chartier [Wed, 22 Apr 2015 20:56:20 +0000 (13:56 -0700)]
Move mirror::ArtMethod to native
Optimizing + quick tests are passing, devices boot.
TODO: Test and fix bugs in mips64.
Saves 16 bytes per most ArtMethod, 7.5MB reduction in system PSS.
Some of the savings are from removal of virtual methods and direct
methods object arrays.
Bug:
19264997
(cherry picked from commit
e401d146407d61eeb99f8d6176b2ac13c4df1e33)
Change-Id: I622469a0cfa0e7082a2119f3d6a9491eb61e3f3d
Fix some ArtMethod related bugs
Added root visiting for runtime methods, not currently required
since the GcRoots in these methods are null.
Added missing GetInterfaceMethodIfProxy in GetMethodLine, fixes
--trace run-tests 005, 044.
Fixed optimizing compiler bug where we used a normal stack location
instead of double on ARM64, this fixes the debuggable tests.
TODO: Fix JDWP tests.
Bug:
19264997
Change-Id: I7c55f69c61d1b45351fd0dc7185ffe5efad82bd3
ART: Fix casts for 64-bit pointers on 32-bit compiler.
Bug:
19264997
Change-Id: Ief45cdd4bae5a43fc8bfdfa7cf744e2c57529457
Fix JDWP tests after ArtMethod change
Fixes Throwable::GetStackDepth for exception event detection after
internal stack trace representation change.
Adds missing ArtMethod::GetInterfaceMethodIfProxy call in case of
proxy method.
Bug:
19264997
Change-Id: I363e293796848c3ec491c963813f62d868da44d2
Fix accidental IMT and root marking regression
Was always using the conflict trampoline. Also included fix for
regression in GC time caused by extra roots. Most of the regression
was IMT.
Fixed bug in DumpGcPerformanceInfo where we would get SIGABRT due to
detached thread.
EvaluateAndApplyChanges:
From ~2500 -> ~1980
GC time: 8.2s -> 7.2s due to 1s less of MarkConcurrentRoots
Bug:
19264997
Change-Id: I4333e80a8268c2ed1284f87f25b9f113d4f2c7e0
Fix bogus image test assert
Previously we were comparing the size of the non moving space to
size of the image file.
Now we properly compare the size of the image space against the size
of the image file.
Bug:
19264997
Change-Id: I7359f1f73ae3df60c5147245935a24431c04808a
[MIPS64] Fix art_quick_invoke_stub argument offsets.
ArtMethod reference's size got bigger, so we need to move other args
and leave enough space for ArtMethod* and 'this' pointer.
This fixes mips64 boot.
Bug:
19264997
Change-Id: I47198d5f39a4caab30b3b77479d5eedaad5006ab
Nicolas Geoffray [Tue, 2 Jun 2015 10:25:09 +0000 (11:25 +0100)]
Fix another source of undeterministic inlining.
Needing an environment and throwing might depend in which
caller first tries to inline. HLoadClass has such logic. Therefore,
do not cache the non-inlineable flag for those cases.
Does not apply to aosp, as this code has been removed.
bug:
20037935
Change-Id: I5e3d36ffb832fb3c3f3c604c79cdbf4f37c749e1
Nicolas Geoffray [Mon, 1 Jun 2015 17:13:34 +0000 (17:13 +0000)]
Merge "Fix another source of undeterministic inlining." into mnc-dev
Andreas Gampe [Mon, 1 Jun 2015 15:28:37 +0000 (15:28 +0000)]
Merge "ART: Fix wrong hard-failure handling in verifier" into mnc-dev
Andreas Gampe [Mon, 1 Jun 2015 15:23:43 +0000 (15:23 +0000)]
Merge changes I3eee8df3,I78cc01fb into mnc-dev
* changes:
Fix test 115-native-bridge for Mips.
ART: Add support for special handlers in sigchainlib
Andreas Gampe [Mon, 1 Jun 2015 15:23:39 +0000 (15:23 +0000)]
Merge "ART: Use v2 version of NativeBridge struct" into mnc-dev
Nicolas Geoffray [Wed, 27 May 2015 11:53:36 +0000 (12:53 +0100)]
Fix another source of undeterministic inlining.
bug:
20037935
(cherry picked from commit
5ae1325048f3d2143f62735ed41acebf28ee6742)
Change-Id: I4bbec76988cfd3061084055c16d2d9ef46845f0a
Nicolas Geoffray [Mon, 1 Jun 2015 07:58:14 +0000 (07:58 +0000)]
Merge "Make inlining deterministic." into mnc-dev
Douglas Leung [Sat, 30 May 2015 02:50:02 +0000 (19:50 -0700)]
Fix test 115-native-bridge for Mips.
(cherry picked from commit
d0af5479d61013362e43ae15cb69c9f8da915e08)
Change-Id: I3eee8df39e774d233f11c43a5578f407bfdebca1
Andreas Gampe [Sat, 23 May 2015 01:31:50 +0000 (18:31 -0700)]
ART: Add support for special handlers in sigchainlib
Add support for a special managed handler in sigchainlib that
will be called as the first user handler.
Use this support for native bridge v2. Extend test 115-native-bridge
to test the functionality.
Bug:
20217701
(cherry picked from commit
03c2cc89428914adf52229d6a3867eef6127911a)
Change-Id: I78cc01fbdabe169154ff6b94c9f3ddb95b5c7448
Andreas Gampe [Thu, 21 May 2015 01:01:30 +0000 (18:01 -0700)]
ART: Use v2 version of NativeBridge struct
Update the struct to the new version, but without implementation.
Bug:
20217701
(cherry picked from commit
540cc3d49c858e670591d8c00fcd87576cdbede6)
Change-Id: Ic23a60b949f119c7d8b0e7cb27a61e6c16532a23
Jeff Hao [Fri, 29 May 2015 17:53:17 +0000 (10:53 -0700)]
Add unstarted runtime test for String.<init>.
Bug:
21173514
Change-Id: I369806644879fffc01bdc123d81a551e6eedf465
Vladimir Marko [Fri, 8 May 2015 11:26:59 +0000 (12:26 +0100)]
Fix double-exception in super-class method validation.
Bug:
19333589
(cherry picked from commit
d5e5a0e61b5b1fe0aaa7edc25a08adebd5a92353)
Change-Id: If394a678ef6271c1119dac3f959773f23e789c56
Andreas Gampe [Fri, 29 May 2015 00:49:39 +0000 (00:49 +0000)]
Merge "ART: Add Mips o32 callee-save registers to SaveAll frame" into mnc-dev
Andreas Gampe [Thu, 28 May 2015 01:31:42 +0000 (18:31 -0700)]
ART: Add Mips o32 callee-save registers to SaveAll frame
The floating point registers f20-f31 are callee-save in the Mips
ABI. While the managed code does not touch them, they need to be
saved when throwing an exception, so that they will be correctly
restored and not smashed.
Bug:
21266656
(cherry picked from commit
a4e0e67611f54180694e0807d6599c1132269b6c)
Change-Id: Ia96d52ce7fb41bf604da1797ce4d7a703e292415
Jeff Hao [Thu, 28 May 2015 21:32:07 +0000 (14:32 -0700)]
Add JNI String.<init> test for global and weak global refs.
Test for change:
https://googleplex-android-review.git.corp.google.com/#/c/698642/
Change-Id: I3e1182d42f9e24448d8616000de58ebfe1e50fb8
Sebastien Hertz [Thu, 28 May 2015 09:00:57 +0000 (11:00 +0200)]
Fix single-step in native thread
If we attempt to single-step in a thread that is not running Java
code, there is no current method on the stack. So we need to check
for null before trying to decode debug info.
Bug:
21320157
(cherry picked from commit
52f5f93873ec244320e05b033243c5c7a3ae40e2)
Change-Id: I961108071dd58a23f6327370b00c35a4cabe0aa7
Jeff Hao [Thu, 28 May 2015 02:46:01 +0000 (02:46 +0000)]
Merge "Fix updating of JNI references for String.<init>." into mnc-dev
Jeff Hao [Thu, 28 May 2015 02:29:29 +0000 (19:29 -0700)]
Fix updating of JNI references for String.<init>.
Was missing updates to globals and weak globals.
Bug:
21288130
Bug:
21440428
Change-Id: I5f801f68b61f6b066b441b92ace367e6ba434789
Dmitriy Ivanov [Wed, 27 May 2015 22:34:38 +0000 (15:34 -0700)]
Fix valgrind includes
external/valgrind structure has been changed
to reflect upstream projects struture.
Making corresponding changes to art includes.
Bug: http://b/
21471495
Change-Id: I2434a270f88502528df82c4820afedb002b65b8d
(cherry picked from commit
6130f739d8f35e01037ba5c3ec5e5a1057900479)
Dmitriy Ivanov [Sat, 23 May 2015 01:03:34 +0000 (18:03 -0700)]
Reset libc targetSdkVersion
This will let libc/linker know which sdk version
is targeted by the app. This is needed to work around
incorrect DT_NEEDED entries in old apps.
Bug: http://b/
21364029
Change-Id: I2f2907104106aac6c8831381ca171dc0302e6a9d
(cherry picked from commit
15efde2badd80b8b124f8e1444165e4afbac17f0)
Andreas Gampe [Tue, 26 May 2015 20:35:39 +0000 (13:35 -0700)]
ART: Fix wrong hard-failure handling in verifier
Correctly return null for the method when a hard failure is
encountered. Also improve logging for order failures. Add a
regression test.
Bug:
20224106
Change-Id: I6e08202617147378b204af169308b67fc69f92c6
Vladimir Marko [Wed, 27 May 2015 13:07:08 +0000 (14:07 +0100)]
ART: Use kBadFprBase on mips64 and arm64.
Fix incorrect usage of kBadGprBase to avoid misleading and
ambiguous data in crash investigations.
Bug:
21266656
(cherry picked from commit
5b09ea0af468d9232cf725ac7f7e73c145892c5c)
Change-Id: Ia4fc60a5d61471c7a6c8fd0545a897ace43d90be
Mathieu Chartier [Wed, 27 May 2015 02:48:16 +0000 (19:48 -0700)]
Add missing oatdump test dependency
(cherry picked from commit
ea216b3abf96d4121a96e8fdab5ce97d7f4d1c2b)
Bug:
18000316
Change-Id: Ic828e8be8e5d3ef7e112facd0db57013740d449a
Mathieu Chartier [Tue, 26 May 2015 21:44:35 +0000 (14:44 -0700)]
Add oatdump test
(cherry picked from commit
19510f02b011e545665f6219e6144c8e47aed5f0)
Bug:
18000316
Change-Id: Ic22d63dc64a9b7a492b5e6bfbd4c75f30c35dfd5
Vladimir Marko [Tue, 26 May 2015 19:42:30 +0000 (20:42 +0100)]
Fix mac build: cast fpr callee save addresses.
Bring back reinterpret_cast<>s removed by
https://android-review.googlesource.com/151309 .
They are needed for mac builds.
Bug :
13925192
(cherry picked from commit
0eb42511405f3919572a480a009426dfb776cfb5)
Change-Id: If6e0cdbadcc84a046abe03d7eca5cccd0a50d141
Vladimir Marko [Tue, 26 May 2015 18:57:30 +0000 (19:57 +0100)]
Fix mac build: include <sys/time.h> in time_utils.cc .
Required for gettimeofday(). Not needed in utils.cc .
Bug:
13925192
(cherry picked from commit
ce39200492e2715a94d96fbd39fa42b46115c255)
Change-Id: If40694965b84605819cac11a7937821aa74db82b
Vladimir Marko [Tue, 19 May 2015 17:08:00 +0000 (18:08 +0100)]
ART: Clean up arm64 kNumberOfXRegisters usage.
Avoid undefined behavior for arm64 stemming from 1u << 32 in
loops with upper bound kNumberOfXRegisters.
Create iterators for enumerating bits in an integer either
from high to low or from low to high and use them for
<arch>Context::FillCalleeSaves() on all architectures.
Refactor runtime/utils.{h,cc} by moving all bit-fiddling
functions to runtime/base/bit_utils.{h,cc} (together with
the new bit iterators) and all time-related functions to
runtime/base/time_utils.{h,cc}. Improve test coverage and
fix some corner cases for the bit-fiddling functions.
Bug:
13925192
(cherry picked from commit
80afd02024d20e60b197d3adfbb43cc303cf29e0)
Change-Id: I905257a21de90b5860ebe1e39563758f721eab82
Sebastien Hertz [Tue, 26 May 2015 09:53:39 +0000 (11:53 +0200)]
JDWP: fix breakpoint on catch statement
Setting a breakpoint on a catch statement in the source actually
installs the breakpoint on a DEX move-exception instruction. At this
point, an exception is pending in the current thread.
The issue is no exception must be pending in the current thread to
report the breakpoint event. This is required to be able to call JNI
functions to create JDWP ids.
This CL fixes it by clearing the pending exception before reporting
event and restore it after reporting the event.
Bug:
21382373
(cherry picked from commit
de48aa6708a3d5dacf7db3d64965e23261fb15d3)
Change-Id: Ie107eb47b6f2559c4fa5297f4033e07baad06f38
Nicolas Geoffray [Tue, 26 May 2015 13:35:06 +0000 (14:35 +0100)]
Make inlining deterministic.
Only the case where two methods are not in the same dex
file could lead to undeterministic behavior.
bug:
20037935
(cherry picked from commit
ff199d84b02efe6cd7162a3c414db99240592454)
Change-Id: If1b0bff3a228be4caec9a068210b1d4d2a7bae1f
Hiroshi Yamauchi [Fri, 22 May 2015 22:32:42 +0000 (22:32 +0000)]
Merge "Fix for potential moving GC bugs around proxy class." into mnc-dev
Hiroshi Yamauchi [Thu, 21 May 2015 19:05:27 +0000 (12:05 -0700)]
Fix for potential moving GC bugs around proxy class.
- Handlerize proxy_class which is live across multiple allocation
points in ClassLinker::CreateProxyClass().
- In ClassLinker::CreateProxyClass(), insert a proxy class into the
class table before creating ArtFields for it (and update it later in
LinkClass()) because the field roots (ArtField::declaring_class_)
won't be updated by GC unless the class is in the class table. If GC
happens before they are updated by FixupTemporaryDeclaringClass()
from LinkClass(), FixupTemporaryDeclaringClass() may not update the
field roots correctly because the old class may already be moved but
the fields roots may not. Reduce a window of time where the fields
roots could be stale.
- In ClassLinker::LinkClass(), directly wrap a new class in a handle
to avoid a window of time where new_class may be potentially stale.
- Print more diagnostic info about the holder of the field upon a mark
sweep invalid ref crash.
- Add an additional sanity check in Field::GetArtField().
Bug:
20557050
Change-Id: I9ad32d304922da96b7e1fad262d97de21cbac776
Andreas Gampe [Thu, 21 May 2015 21:34:05 +0000 (21:34 +0000)]
Merge "ART: Sometimes even empty methods take forever to verify" into mnc-dev
Andreas Gampe [Thu, 21 May 2015 21:15:37 +0000 (21:15 +0000)]
Merge "Revert "DO NOT MERGE ART: Turn off duplicate-classes checking"" into mnc-dev
Andreas Gampe [Thu, 21 May 2015 21:06:46 +0000 (14:06 -0700)]
ART: Sometimes even empty methods take forever to verify
In cases of very high load and/or bad scheduling, the verifier may
take longer than the threshold duration to verify an empty method.
The LargeMethod detection needs to accept that the code_item may
be null.
Bug:
21364300
Change-Id: Iceff3e4688cc1a5fe7a836f7a9bf6c49a392b618
Andreas Gampe [Thu, 21 May 2015 20:33:40 +0000 (20:33 +0000)]
Revert "DO NOT MERGE ART: Turn off duplicate-classes checking"
This reverts commit
aafcfca5fe545365ef377fff2897b8a908f03e71.
Bug:
21333911
Change-Id: Ibbc6b1be8bebc1d0ab9c5955b34eb89f40577baa
Andreas Gampe [Sat, 16 May 2015 02:24:12 +0000 (19:24 -0700)]
ART: Refactor UnstartedRuntime for testing
Expose the UnstartedRuntime implementation functions as private static
methods of a class. Add a gtest that can invoke these functions. Add
sample tests for String and Memory.
Bug:
21173514
(cherry picked from commit
799681b176ad25437ce2849639f54f610dcbf684)
Change-Id: Ib5bde6347fafaf7607c642542ea7d5938ff4b1df
Andreas Gampe [Thu, 21 May 2015 15:49:28 +0000 (15:49 +0000)]
Merge "DO NOT MERGE ART: Turn off duplicate-classes checking" into mnc-dev
Jeff Hao [Thu, 21 May 2015 01:39:36 +0000 (01:39 +0000)]
Merge "Intercept JNI invocation of String.<init> methods." into mnc-dev
Jeff Hao [Wed, 20 May 2015 03:30:23 +0000 (20:30 -0700)]
Intercept JNI invocation of String.<init> methods.
libmono uses JNI AllocObject and CallNonvirtualVoidMethod to create and
initialize a string instead of using the recommended NewObject. This
change adds an intercept to change the String.<init> call to a
StringFactory call instead. Then, it uses the object id of the original
string object referrer and maps it to the result of the StringFactory.
Bug:
21288130
Change-Id: Ib4db402c178bc37188d5c5faf30b6e4fdc747b17
Andreas Gampe [Thu, 21 May 2015 01:15:36 +0000 (01:15 +0000)]
Merge "ART: Fix RegisterNative order" into mnc-dev
Andreas Gampe [Wed, 20 May 2015 23:09:36 +0000 (16:09 -0700)]
ART: Fix gtest dependency
Gtests should depend on the debug version of libart-disassembler.
Bug:
21273194
(cherry picked from commit
0ae3bf872d3898ed6826216f3f8a04c819e1426b)
Change-Id: Iaf216a1c0cff650296ebf83fe4e8352f5614cb28
Hiroshi Yamauchi [Wed, 20 May 2015 23:10:39 +0000 (23:10 +0000)]
Merge "Fix a cts crash around proxy class fields." into mnc-dev
Hiroshi Yamauchi [Wed, 20 May 2015 22:51:29 +0000 (15:51 -0700)]
Fix a cts crash around proxy class fields.
org.apache.harmony.tests.java.io.SerializationStressTest4#test_writeObject_Proxy
As the static fields of proxy classes share the dex file indices, they
shouldn't be resolved in the dex cache or else Field::GetArtField()
may return a wrong art field that belong to a different proxy class.
Bug:
20557050
Change-Id: If672c0e67bc49e672e34d75ffbe29c65f5a423b9
Andreas Gampe [Wed, 20 May 2015 21:48:10 +0000 (14:48 -0700)]
DO NOT MERGE ART: Turn off duplicate-classes checking
Bug:
21333911
Change-Id: I9832526a6d2996aadb459e3ec0930cbaf151478a
Andreas Gampe [Mon, 18 May 2015 22:52:22 +0000 (15:52 -0700)]
ART: Fix RegisterNative order
First check for both direct and virtual methods in the current class,
then move to the parent.
Optimize registration by checking first whether the current method
under test is native. This slows down registering implementations
in parent classes. Add a CheckJNI warning for this.
Add a run-test to check the behavior. Fix host comparison testing.
Bug:
19569721
(cherry picked from commit
3f1dc56914177993b1b018bf21ce7d39d7feecda)
Change-Id: I61e77117d96310632aad123d7f1279d0f834dc99
Sebastien Hertz [Wed, 20 May 2015 06:55:21 +0000 (06:55 +0000)]
Merge "Fix debuggable compiler flag detection for secondary dex files" into mnc-dev
Andreas Gampe [Tue, 19 May 2015 03:50:12 +0000 (20:50 -0700)]
ART: Change huge-method warning in verifier
Change the logging to be included in the took-a-long-time warning,
instead of being a non-descriptive standalone. This should reduce
the noise.
Bug:
21149450
(cherry picked from commit
a4f5bf6ea99d6b3095a4ca261352729633bc44c5)
Change-Id: Ia4493dd12f54a8452a46541f6da86530584016ea
Sebastien Hertz [Wed, 13 May 2015 10:14:05 +0000 (12:14 +0200)]
Fix debuggable compiler flag detection for secondary dex files
Compiles secondary dex files like the primary dex file: if it has
been compiled with the --debuggable flag, compile secondary dex files
with the --debuggable flag too.
Therefore, dex files loaded at runtime are compiled the same way as
dex files compiled at install time on the classpath (excluding the
boot image that is not compiled debuggable).
Also adds debuggable key in the oat header and bump the oat version.
Bug:
20944228
(cherry picked from commit
0de1133ba600f299b3d67938f650720d9f859eb2)
Change-Id: If6b2236e7fe547cc421f57b573043748018d3ae0
Richard Uhler [Mon, 18 May 2015 15:18:55 +0000 (15:18 +0000)]
Merge "Test and fix bug checking status of secondary multidex files." into mnc-dev
Richard Uhler [Thu, 14 May 2015 20:21:13 +0000 (13:21 -0700)]
Test and fix bug checking status of secondary multidex files.
We were accidentally returning 'false' instead of 'true' to indicate a native
multidex file with an up-to-date primary dex file checksum but out-of-date
secondary dex file checksum is out of date.
Bug:
21022815
Change-Id: I43d360ca9251d0df38ad7f8289f6ef339c0a3280
Sebastien Hertz [Wed, 8 Apr 2015 07:36:07 +0000 (09:36 +0200)]
JDWP: more GC safety
Ensures GC safety when keeping references that may be moved by GC:
- SingleStepControl: stores ArtMethod* in a GcRoot
- ModBasket: stores references in a StackHandleScope
Bug:
18166750
(cherry picked from commit
261bc044a3575512869586593e99e97cd8b1c321)
Change-Id: I35971a901537956739d1f089d61cb4ea9dc6c93d
Chih-Hung Hsieh [Fri, 15 May 2015 23:09:19 +0000 (23:09 +0000)]
Merge "Undo arm64 workaround." into mnc-dev
Chih-Hung Hsieh [Tue, 12 May 2015 19:06:57 +0000 (12:06 -0700)]
Undo arm64 workaround.
Now we have new llvm to compile art for arm64 without -fno-vectorize.
BUG:
21080674
Change-Id: Id1b3a88ab9d4cf78b7f0dfd8cb435ed6e7ab7cb2
(cherry picked from commit
b6cee946d44dc1e961c9656830b567e38104f896)
Hiroshi Yamauchi [Fri, 15 May 2015 22:47:18 +0000 (22:47 +0000)]
Merge "Print more info on mark sweep invalid ref crash." into mnc-dev
Hiroshi Yamauchi [Fri, 15 May 2015 22:23:30 +0000 (15:23 -0700)]
Print more info on mark sweep invalid ref crash.
- Add extra info about the field holder.
- Suspend threads when verifying roots.
Bug:
20557050
Change-Id: Ia1f2269aaa3b1a81d0594e781e439a5decfb82f9
Kenny Root [Thu, 14 May 2015 22:58:17 +0000 (15:58 -0700)]
Add new style String operations during compilation init, part 2
This allows some class initializers that deal with Strings to run during
compilation. However, java.lang.Character.toUpperCase() is blocking
initialization for some parts.
(cherry picked from commit
57f91e8956f4496391bff028a1b990540cc91c22)
Bug:
21036900
Change-Id: Ia969c6a643f510bc798f94fb10bca1a68fe2ae67
Elliott Hughes [Fri, 15 May 2015 15:25:58 +0000 (08:25 -0700)]
Stop wrapping close(2) with TEMP_FAILURE_RETRY in art.
Note that all the TEMP_FAILURE_RETRY([A-Z] calls in this file are
wrong. The member functions return -errno, so only an EPERM failure
will return -1 (but errno should be EPERM, so you won't loop).
Bug: http://b/
20501816
Change-Id: Ibde6a2c7bf92469d008d6d510f9391669ae37a84
(cherry picked from commit
6a887d6a1d1ca089b48663a5ae11f973b98785a9)