OSDN Git Service
Andy McFadden [Tue, 20 Apr 2010 20:34:11 +0000 (13:34 -0700)]
Merge "Make plain "-Xint" work." into dalvik-dev
Andy McFadden [Tue, 20 Apr 2010 20:17:52 +0000 (13:17 -0700)]
Make plain "-Xint" work.
Change-Id: I7a09a886df9722b60a861969f3e1a1a45988068a
Brian Carlstrom [Tue, 20 Apr 2010 16:18:09 +0000 (09:18 -0700)]
Merge "Merge commit '
a0a3ad10' into manualmerge" into dalvik-dev
Carl Shapiro [Tue, 20 Apr 2010 03:10:35 +0000 (20:10 -0700)]
Since we no longer make use of formatted I/O in the unlock monitor
exception case, revert to using dvmThrowException which does not treat
the description string as an output format.
Change-Id: Ice1dc04cb6d1e887af718022cc0b03c6d90f9a19
Brian Carlstrom [Tue, 20 Apr 2010 00:30:47 +0000 (17:30 -0700)]
Merge commit '
a0a3ad10' into manualmerge
Conflicts:
libcore/x-net/src/main/native/org_apache_harmony_xnet_provider_jsse_NativeCrypto.cpp
Change-Id: I23508e348c5de4103e9dda619ca230d0575f57ca
Carl Shapiro [Tue, 20 Apr 2010 00:01:40 +0000 (17:01 -0700)]
am
5005a521: am
612fa566: am
af69cf8d: Fix several issues with the lock sampling code. This change obsoletes the lockProfSample flag.
Merge commit '
5005a5210b941d18b7742469ba622b1c8a754b1e' into dalvik-dev
* commit '
5005a5210b941d18b7742469ba622b1c8a754b1e':
Fix several issues with the lock sampling code. This change obsoletes
Carl Shapiro [Tue, 20 Apr 2010 00:01:38 +0000 (17:01 -0700)]
am
bd4eaf95: am
54c9323e: am
6b4ba58e: Add a mechanism to long a sample of long running lock acquisitions.
Merge commit '
bd4eaf959e21f3e048366141896ce262f22e0e2c' into dalvik-dev
* commit '
bd4eaf959e21f3e048366141896ce262f22e0e2c':
Add a mechanism to long a sample of long running lock acquisitions.
Jesse Wilson [Tue, 20 Apr 2010 00:01:34 +0000 (17:01 -0700)]
am
a731357c: (-s ours) am
a8984fe8: (-s ours) am
ddc0d4c2: (-s ours) DO NOT MERGE. Suppressing tests that fail in CTS\' test runner.
Merge commit '
a731357c0cadaca3863f6fd0de1b8ffb39464220' into dalvik-dev
* commit '
a731357c0cadaca3863f6fd0de1b8ffb39464220':
DO NOT MERGE. Suppressing tests that fail in CTS' test runner.
Elliott Hughes [Mon, 19 Apr 2010 22:28:23 +0000 (15:28 -0700)]
Fix U_FILE_ACCESS_ERROR RuntimeException in String.getBytes("GB18030").
Bug:
2606807
Change-Id: Ie639ce6380e42e0eed2a7e360504f86a22f1fed8
Elliott Hughes [Mon, 19 Apr 2010 21:29:52 +0000 (14:29 -0700)]
Fix several BigInteger failures.
From Java 7 on, BigInteger also accepts a leading "+" in its string
constructors.
BigInteger has always claimed to accept non-ASCII digits, but our
implementation never has.
BigInteger.isProbablePrime is defined to return true for certainty <= 0,
but OpenSSL (on which we're based) takes the opposite stance.
Change-Id: I00bfc591a4d583460f71b7eec3de91bf6b03cd87
Elliott Hughes [Mon, 19 Apr 2010 21:11:56 +0000 (14:11 -0700)]
Merge "Speed up %d for locales with non-ASCII digits." into dalvik-dev
Elliott Hughes [Mon, 19 Apr 2010 18:10:41 +0000 (11:10 -0700)]
Speed up %d for locales with non-ASCII digits.
The small "localizeDigits" part of this change is the big speedup. It takes
%d in the "ar" Locale from 5x more expensive than "en_US", to just a small
percentage more expensive.
The big part of this change: removing the useless Transformer and FloatUtils
classes touches many lines (mostly just changing the level of indentation),
and only has a small percentage effect on performance. The motivation for
this is actually primarily that it makes the code easier to read, and makes
it easier to share code.
Change-Id: I07e86a3ee448510d9ae36579a9af283349f777f1
Brian Carlstrom [Mon, 19 Apr 2010 20:58:57 +0000 (13:58 -0700)]
Merge "openssl-1.0.0 upgrade"
Jesse Wilson [Mon, 19 Apr 2010 02:55:45 +0000 (19:55 -0700)]
Tidying up the error reporting.
We were spamming our logs with a bunch of redundant messages,
particularly surrounding timeouts. This should reduce the log
output to only the useful stuff when a test takes too long
to complete.
Also not blowing away the full XML report when a few tests
get lost along the way.
Elliott Hughes [Sun, 18 Apr 2010 17:38:03 +0000 (10:38 -0700)]
Remove a dangling import, and another unused Messages.java.
Also tidy up our two awt.font classes.
Change-Id: I931d348eef7265190f76c31f74c9a5d9ec5d6597
Elliott Hughes [Sun, 18 Apr 2010 17:17:46 +0000 (10:17 -0700)]
Merge "Remove the nio Messages.java I missed, and the three now-unused imports." into dalvik-dev
Elliott Hughes [Sun, 18 Apr 2010 17:00:54 +0000 (10:00 -0700)]
Remove the nio Messages.java I missed, and the three now-unused imports.
I seem to be having trouble deleting files lately...
Change-Id: I019acd19aa61acdeeee1ab6f7fb8a063a41a5534
Jesse Wilson [Sat, 17 Apr 2010 22:05:08 +0000 (15:05 -0700)]
Small vogar changes in response to the review for change 49052.
Brian Carlstrom [Sat, 17 Apr 2010 18:45:39 +0000 (11:45 -0700)]
Merge "Teaching vogar to run prebuilt .jar files." into dalvik-dev
Elliott Hughes [Sat, 17 Apr 2010 16:14:48 +0000 (09:14 -0700)]
Merge "Remove "messages" from the logging, math, and nio_char modules." into dalvik-dev
Elliott Hughes [Sat, 17 Apr 2010 06:49:40 +0000 (23:49 -0700)]
Remove two dead files.
java_lang_Character.cpp has long been dead; I killed AndroidSystemNatives.h
this afternoon. I'm not sure how their "git rm"s didn't get included in my
earlier patch.
Change-Id: Iaa73935846a474dbd25792ed0176ccea092e9f7c
Elliott Hughes [Sat, 17 Apr 2010 01:08:47 +0000 (18:08 -0700)]
Remove "messages" from the logging, math, and nio_char modules.
Change-Id: Ib61b132ce17fdd37956889e855bda35956f8ae62
Jesse Wilson [Sat, 17 Apr 2010 02:33:37 +0000 (19:33 -0700)]
Teaching vogar to run prebuilt .jar files.
Also trying to simplify some internals. We always build to
intermediate .jar files now.
I've renamed TestFinder to RunnerSpec and fixed its users.
Elliott Hughes [Sat, 17 Apr 2010 01:32:04 +0000 (18:32 -0700)]
Merge "Move the libcore registration out of libnativehelpers and into libcore." into dalvik-dev
Carl Shapiro [Sat, 17 Apr 2010 01:27:51 +0000 (18:27 -0700)]
am
612fa566: am
af69cf8d: Fix several issues with the lock sampling code. This change obsoletes the lockProfSample flag.
Carl Shapiro [Sat, 17 Apr 2010 01:19:00 +0000 (18:19 -0700)]
am
af69cf8d: Fix several issues with the lock sampling code. This change obsoletes the lockProfSample flag.
Merge commit '
af69cf8d062a42300852ddee6bcb9f189f5a1b83' into kraken
* commit '
af69cf8d062a42300852ddee6bcb9f189f5a1b83':
Fix several issues with the lock sampling code. This change obsoletes
Carl Shapiro [Sat, 17 Apr 2010 01:11:44 +0000 (18:11 -0700)]
Merge "Refinements to array and class object size computation." into dalvik-dev
Carl Shapiro [Sat, 17 Apr 2010 00:55:27 +0000 (17:55 -0700)]
Refinements to array and class object size computation.
* Rename dvmArrayObjectLength to dvmArrayObjectSize. In the context
of arrays length may refer to the length of the array data and not
the size of the array object instance.
* Add a new method dvmClassObjectSize which computes the total size of
a class object including its static fields.
Change-Id: I693ec8e66dfa5df35c9f35474c80411ea917c899
Elliott Hughes [Sat, 17 Apr 2010 00:44:12 +0000 (17:44 -0700)]
Move the libcore registration out of libnativehelpers and into libcore.
Bug: 754114
Change-Id: Iaa03def509c10cbaa12fd2128584b93d4be4a6b7
Carl Shapiro [Sat, 17 Apr 2010 00:33:15 +0000 (17:33 -0700)]
Fix several issues with the lock sampling code. This change obsoletes
the lockProfSample flag.
Change-Id: Idb1846fc3db4ff73678a0dae46aabbb5dbb8b957
Carl Shapiro [Sat, 17 Apr 2010 00:31:59 +0000 (17:31 -0700)]
Merge "Remove an unused enum." into dalvik-dev
Carl Shapiro [Fri, 16 Apr 2010 22:44:09 +0000 (15:44 -0700)]
Remove an unused enum.
Change-Id: I792a5a6b890ac37b3e1fa1f3e4a6413b7621d115
Elliott Hughes [Fri, 16 Apr 2010 21:14:28 +0000 (14:14 -0700)]
Merge LocaleData and Resources, rename Resources to ICU.
Also move our ICU tests into our little tree of tests.
Bug:
2596471
Change-Id: I73b53d74c26ef9bf670f12cac58b51ba61eefead
Brian Carlstrom [Fri, 16 Apr 2010 18:55:48 +0000 (11:55 -0700)]
merging FRF15 for dalvik-dev
Change-Id: Ic8c9dea280d38cba8e55f962f0d8df843aea8e35
Jesse Wilson [Fri, 16 Apr 2010 18:32:30 +0000 (11:32 -0700)]
Merge "Fixing TreeMap serialization issues." into dalvik-dev
Elliott Hughes [Fri, 16 Apr 2010 18:06:58 +0000 (11:06 -0700)]
Change DecimalFormatSymbols to have a field per symbol.
I'd been wanting to do this for some time, but cleaning up the recent
performance changes I made to Formatter was the final straw.
Change-Id: I6d516de66a0bed5e759bca590b4cc124ce2eb712
Jesse Wilson [Fri, 16 Apr 2010 00:32:27 +0000 (17:32 -0700)]
Fixing TreeMap serialization issues.
When I was testing this on the JDK I had my own TreeMap on
the classpath, which caused my tests to pass even when they
shouldn't have.
The new serialization data was generated by the JDK's
collections, both v5 and v6. The tests pass on both Dalvik
and the JDK.
Carl Shapiro [Thu, 15 Apr 2010 23:59:59 +0000 (16:59 -0700)]
am
54c9323e: am
6b4ba58e: Add a mechanism to long a sample of long running lock acquisitions.
Elliott Hughes [Thu, 15 Apr 2010 23:31:51 +0000 (16:31 -0700)]
Merge "java.util.Formatter: fix localization of numbers." into dalvik-dev
Elliott Hughes [Thu, 15 Apr 2010 21:16:58 +0000 (14:16 -0700)]
java.util.Formatter: fix localization of numbers.
Bug:
2301911
Change-Id: Ic594702a80bdef2e8fd97e52fd5fa301def19cac
Brian Carlstrom [Thu, 15 Apr 2010 21:00:03 +0000 (14:00 -0700)]
openssl-1.0.0 upgrade
external/openssl
Updated version to 1.0.0
openssl.version
Updated small records patch for 1.0.0. This is probably the most significant change.
patches/small_records.patch
Removed bad_version.patch since fix is included in 0.9.8n and beyond
patches/README
patches/bad_version.patch
openssl.config
Changed import_openssl.sh to generate armv4 asm with the 1.0.0
scripts, not our backported 0.9.9-dev backported version in
patches/arm-asm.patch.
import_openssl.sh
openssl.config
patches/README
patches/arm-asm.patch
Added -DOPENSSL_NO_STORE to match ./Configure output
Added -DOPENSSL_NO_WHIRLPOOL (no-whrlpool) to skip new optional cipher
android-config.mk
openssl.config
Fixed import to remove include directory during import like other
imported directories (apps, ssl, crypto)
import_openssl.sh
Updated UNNEEDED_SOURCES. Pruned Makefiles which we don't use.
openssl.config
Updated to build newly required files
patches/apps_Android.mk
patches/crypto_Android.mk
Disable some new openssl tools
patches/progs.patch
Automatically imported
android.testssl/
apps/
crypto/
e_os.h
e_os2.h
include/
ssl/
dalvik
Change makeCipherList to skip SSLv2 ciphers that 1.0.0 now returns
so there are not duplicate ciphersuite names in getEnabledCipherSuites.
libcore/x-net/src/main/native/org_apache_harmony_xnet_provider_jsse_NativeCrypto.cpp
Updated OpenSSLSocketImpl_cipherauthenticationmethod for new
SSL_CIPHER algorithms -> algorithm_auth (and const-ness)
libcore/x-net/src/main/native/org_apache_harmony_xnet_provider_jsse_NativeCrypto.cpp
Update to const SSL_CIPHER in OpenSSLSessionImpl_getCipherSuite (and cipherauthenticationmethod)
libcore/x-net/src/main/native/org_apache_harmony_xnet_provider_jsse_NativeCrypto.cpp
test_EnabledCipherSuites on both SSLSocketTest and
SSLServerSocketTest caught the makeCipherList problem. However the
asserts where a bit out of sync and didn't give good messages
because they didn't actually show what was going on. As part of
debugging the issue they found, I tried to make align the asserts
and improve their output for the future.
libcore/x-net/src/test/java/tests/api/javax/net/ssl/SSLServerSocketTest.java
libcore/x-net/src/test/java/tests/api/javax/net/ssl/SSLSocketTest.java
vendor/google
Add const to X509V3_EXT_METHOD* for 1.0.0 compatibility
libraries/libjingle/talk/base/openssladapter.cc
Change-Id: I9e848c79772211d3956f8561ec526339b30e24a4
Carl Shapiro [Thu, 15 Apr 2010 22:32:07 +0000 (15:32 -0700)]
am
6b4ba58e: Add a mechanism to long a sample of long running lock acquisitions.
Merge commit '
6b4ba58ec937bfacba626112c46ebd003efbed21' into kraken
* commit '
6b4ba58ec937bfacba626112c46ebd003efbed21':
Add a mechanism to long a sample of long running lock acquisitions.
Jesse Wilson [Thu, 15 Apr 2010 21:51:44 +0000 (14:51 -0700)]
Merge "From scratch implementation of a Navigable TreeMap." into dalvik-dev
Jesse Wilson [Thu, 15 Apr 2010 21:26:42 +0000 (14:26 -0700)]
Merge commit '
da18d995' into manualmerge
Conflicts:
libcore/luni/src/test/java/tests/api/java/io/PipedInputStreamTest.java
Raphael [Thu, 15 Apr 2010 21:07:15 +0000 (14:07 -0700)]
am
d9b87a91: Fix dexdump to build in Windows SDK under Linux
Merge commit '
d9b87a918dfe98793faa08cb59012f652b4b8561' into dalvik-dev
* commit '
d9b87a918dfe98793faa08cb59012f652b4b8561':
Fix dexdump to build in Windows SDK under Linux
Jesse Wilson [Tue, 6 Apr 2010 06:27:51 +0000 (23:27 -0700)]
From scratch implementation of a Navigable TreeMap.
Between Java 5 and Java 6, Harmony's implementation ballooned
from ~2400 lines to ~5900 lines. This implementation is a more
compact ~1700 lines.
This implementation and its views have been rigorously tested
using Google Collections' test framework in addition to the
tests of our own suite.
Jesse Wilson [Thu, 15 Apr 2010 19:34:29 +0000 (12:34 -0700)]
am
a8984fe8: (-s ours) am
ddc0d4c2: (-s ours) DO NOT MERGE. Suppressing tests that fail in CTS\' test runner.
Jesse Wilson [Thu, 15 Apr 2010 13:36:55 +0000 (06:36 -0700)]
am
e09ba122: Fixing PipedInputStreamTest to fill the buffer before dying on a dead reader.
Carl Shapiro [Fri, 9 Apr 2010 22:03:33 +0000 (15:03 -0700)]
Add a mechanism to long a sample of long running lock acquisitions.
Raphael [Tue, 13 Apr 2010 22:19:21 +0000 (15:19 -0700)]
Fix dexdump to build in Windows SDK under Linux
Change-Id: I0e57c81dcfc53d7065d420129d75ac32cda040e2
Jesse Wilson [Thu, 15 Apr 2010 00:08:44 +0000 (17:08 -0700)]
am
ddc0d4c2: (-s ours) DO NOT MERGE. Suppressing tests that fail in CTS\' test runner.
Merge commit '
ddc0d4c2f8d3fea4ddaf6e611ffd73a1169447e8' into kraken
* commit '
ddc0d4c2f8d3fea4ddaf6e611ffd73a1169447e8':
DO NOT MERGE. Suppressing tests that fail in CTS' test runner.
Jesse Wilson [Wed, 14 Apr 2010 21:43:12 +0000 (14:43 -0700)]
DO NOT MERGE. Suppressing tests that fail in CTS' test runner.
These tests have been verified to work successfully in run-core-tests.
We plan on unifying CTS and Dalvik's test running infrastructure in this
quarter so that we don't need such unfortunate hacks.
Test failures: http://b/issue?id=
2564250
Fix test runner: http://b/issue?id=
2215226
Change-Id: I3df2f4de10705dc470e397da2d0f014141573da1
Elliott Hughes [Wed, 14 Apr 2010 19:31:23 +0000 (12:31 -0700)]
Merge "Remove "Messages" from the nio, prefs, sql, text, and x-net modules." into dalvik-dev
Elliott Hughes [Tue, 13 Apr 2010 23:05:10 +0000 (16:05 -0700)]
Remove "Messages" from the nio, prefs, sql, text, and x-net modules.
Also remove an file of unused messages from luni, and inline one of luni's
other files. (There are plenty more.)
Also remove some German translations of bouncycastle messages (that are
actually in English anyway).
Change-Id: I9c565f6f2201a5d877eba5bf0af4ffad7b769984
Jesse Wilson [Wed, 14 Apr 2010 01:02:47 +0000 (18:02 -0700)]
Fixing PipedInputStreamTest to fill the buffer before dying on a dead reader.
Change-Id: Ie34a8bec7b2a34d2933225d6d4e04e0643dbcf99
Carl Shapiro [Tue, 13 Apr 2010 23:48:27 +0000 (16:48 -0700)]
Fix a typo and unbreak the build.
Change-Id: Ib05e4d7602f3614e9eb792bd29ff7288de574e84
Carl Shapiro [Tue, 13 Apr 2010 23:34:33 +0000 (16:34 -0700)]
resolved conflicts for merge of
f0c514ce to dalvik-dev
Change-Id: Ie9274ee7a4e767e197f53bd350c835fedbca3ae9
Elliott Hughes [Tue, 13 Apr 2010 22:30:18 +0000 (15:30 -0700)]
Use the same documentation in all the methods that take a format string.
Change-Id: I8e7d06df72a7b8db9edd17aa2748800329e837fa
Elliott Hughes [Tue, 13 Apr 2010 22:23:01 +0000 (15:23 -0700)]
Remove "Messages" from the annotation module.
Change-Id: I6253fa192bb45fc06db99937aca30ee4dd0bb201
Elliott Hughes [Tue, 13 Apr 2010 21:28:42 +0000 (14:28 -0700)]
Improve documentation about the user's default locale.
Add a bit of text to Locale's class documentation, and include a link from
every method that uses Locale.getDefault(). Also try to consistently say
"user's default locale", as a subliminal hint that this isn't necessarily
the developer's locale, nor en_US.
Bug:
2593000
Change-Id: Ieebe864ed6b9fecbfef5d5415269299739cedd1b
Carl Shapiro [Fri, 9 Apr 2010 22:03:33 +0000 (15:03 -0700)]
Add a mechanism to long a sample of long running lock acquisitions.
Change-Id: Ic91c5f213bcfd72c6f630ef5102da2c6c7cc8ce4
Elliott Hughes [Tue, 13 Apr 2010 01:32:50 +0000 (18:32 -0700)]
Fix String.toLowerCase and toUpperCase.
Rather than try to cope with Lithuanian, let's just hand that one to ICU4C.
I've removed my hand-crafted Azeri/Turkish lowercasing too, in favor of ICU.
Presence of a high surrogate (which implies a supplemental character) is a
good reason to hand over to ICU too.
On the uppercasing side, I've kept our existing hard-coded table and just
added code to defer to ICU for Azeri, Lithuanian, and Turkish (plus
supplemental characters). I don't like the tables, but I don't have proof
that they're incorrect.
Bug:
2340628
Change-Id: I36b556b0444623a5aacc1afc58ebb4d84211d3dc
Elliott Hughes [Mon, 12 Apr 2010 21:29:37 +0000 (14:29 -0700)]
Fix supplementary character support.
Fixes all known bugs in our handling of supplementary characters. This change
introduces a performance regression on the assumption that it won't be released
without a corresponding JIT change to enable the code to be inlined back to
pretty much what it used to be.
Bug:
2587122
Change-Id: I3449c9718bbe32ebe53b6c10454ae1dc82105b59
Elliott Hughes [Mon, 12 Apr 2010 17:29:14 +0000 (10:29 -0700)]
Merge "Throw the same exceptions as the RI from String methods." into dalvik-dev
Brian Carlstrom [Sat, 10 Apr 2010 23:07:24 +0000 (16:07 -0700)]
Merge commit '
7dbf57f6' into manualmerge
Conflicts:
libcore/x-net/src/main/native/org_apache_harmony_xnet_provider_jsse_NativeCrypto.cpp
Change-Id: I7cf63f6a1ca7c7604301cb8bb947633ad704a2c0
Elliott Hughes [Sat, 10 Apr 2010 19:01:10 +0000 (12:01 -0700)]
Throw the same exceptions as the RI from String methods.
String has its own StringIndexOutOfBoundsException subclass of
IndexOutOfBoundsException. We can run more tests if we behave
the same.
The RI only admits to IndexOutOfBoundsException, though,
so our documentation and throws clause shouldn't change.
Change-Id: Ib87777f8a42d4bcac21e8f8f00f4dcbc0ada4201
Brian Carlstrom [Sat, 10 Apr 2010 18:54:58 +0000 (11:54 -0700)]
am
5fdfbeff: Fix SSL_shutdown error reporting to print a proper error instead of "Ok"
Elliott Hughes [Sat, 10 Apr 2010 17:22:05 +0000 (10:22 -0700)]
Merge "Fix build by turning a comment into a doc comment. Clarify logic." into dalvik-dev
Elliott Hughes [Sat, 10 Apr 2010 06:36:48 +0000 (23:36 -0700)]
Fix build by turning a comment into a doc comment. Clarify logic.
Now I see what jessewilson was trying to say earlier; rewrite the post-loop
code to say what I guess it was aiming at, which is avoiding an add followed
by a remove in the not completely uncommon case where limit == 0 and the
regular expression is a terminator rather than a true separator ('\n', say).
(Normally the regular expression is a true separator so there will be no
trailing empty string even though limit is 0: "a,b,c".split(","), for example.)
Also, String.isEmpty is our fasest way of recognizing the empty string;
certainly better than equals("").
Change-Id: I82f4ec49fa58efc178e342cf55d4dfbbdad01c75
Elliott Hughes [Sat, 10 Apr 2010 01:22:02 +0000 (18:22 -0700)]
Merge "Make String.split 10x faster." into dalvik-dev
Brian Carlstrom [Sat, 10 Apr 2010 00:49:11 +0000 (17:49 -0700)]
Fix SSL_shutdown error reporting to print a proper error instead of "Ok"
There are other problematic error reporting cases to be fixed later, but this one seems to be the frequently occuring.
Change-Id: Ia5910cffb60cc694066be03d43ac1bb7eab47357
Jesse Wilson [Fri, 9 Apr 2010 23:48:34 +0000 (16:48 -0700)]
Merge "Latest java.util.concurrent from the JSR 166 project." into dalvik-dev
Elliott Hughes [Fri, 9 Apr 2010 23:31:13 +0000 (16:31 -0700)]
Make String.split 10x faster.
Almost all uses of String.split in the Android codebase use trivial single
literal character separators. This patch optimizes that case to avoid the
use of regular expressions entirely.
The 10x speedup isn't the whole story, because the speedup is really
proportional to the number of separators in the input. 10x is easily
achievable, but the speedup could be arbitrarily high.
Before:
benchmark us logarithmic runtime
PatternSplitComma 84.8 XXXXXXXXXXXXXX||||||||||||||
PatternSplitLiteralDot 85.0 XXXXXXXXXXXXXX||||||||||||||
StringSplitComma 166.3 XXXXXXXXXXXXXXXXXXXXXXXXXXXX|
StringSplitHard 173.6 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
StringSplitLiteralDot 167.7 XXXXXXXXXXXXXXXXXXXXXXXXXXXX|
After:
benchmark us logarithmic runtime
PatternSplitComma 18.9 XXX|||||||||||||||||||||
PatternSplitLiteralDot 19.0 XXX|||||||||||||||||||||
StringSplitComma 18.8 XXX|||||||||||||||||||||
StringSplitHard 174.2 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
StringSplitLiteralDot 18.8 XXX|||||||||||||||||||||
(The benchmarks starting "Pattern" use a precompiled Pattern for performance.
Those starting "String" use String.split and would traditional entail a
temporary Pattern. As you can see, creating Patterns is very expensive for
us, and each one throws a finalizer spanner in the GC works too. The new
fast path avoids all this. I'll commit the benchmark -- along with all the
others I've ever used -- to http://code.google.com/p/dalvik this afternoon.)
Tests? We actually pass _more_ tests after this patch, because the increase
in performance means we don't hit timeouts.
Change-Id: I404298e21a78d72cf5ce6ea675844bf251e3825b
Brian Carlstrom [Fri, 9 Apr 2010 21:22:59 +0000 (14:22 -0700)]
Disable our new javax.net.ssl tests until hangs are resolved
While the hangs all seem to happen during handshaking, there is more
than one root cause. Until I have things working reliably, disabling
the tests to avoid fouling up the continuous build.
Change-Id: Ia57cac2e49284a1050a72d9ea77813307eff5ea8
Jesse Wilson [Wed, 7 Apr 2010 22:07:54 +0000 (15:07 -0700)]
Latest java.util.concurrent from the JSR 166 project.
Code was downloaded from CVS on 2010-april-7 at 10:00pst
http://gee.cs.oswego.edu/cgi-bin/viewcvs.cgi/jsr166/src/main/java/util/concurrent/
The new interfaces and ArrayDeque class are all from Harmony.
Elliott Hughes [Fri, 9 Apr 2010 17:39:25 +0000 (10:39 -0700)]
Merge "An InlineNative for String.isEmpty, so it's not slower than length() == 0." into dalvik-dev
Andy McFadden [Fri, 9 Apr 2010 14:32:42 +0000 (07:32 -0700)]
Merge "Move the furniture around some more." into dalvik-dev
Andy McFadden [Wed, 7 Apr 2010 22:56:16 +0000 (15:56 -0700)]
Move the furniture around some more.
Mostly just moving things around, with minor changes to behavior.
- Instead of walking through all classes twice (once for verification,
once for optimization), we now walk through them once and do both
operations on a given class before moving on to the next.
- If verification and optimization were disabled, the VM used a special
"no fork + exec" path. It adds complexity for little benefit, so
it's gone.
- Reduced the amount of stuff being passed as arguments through multiple
layers of functions. Notably, a pointer to a read-only lookup table
is now accessed via a global.
- The PROFILE_FIELD_ACCESS define now just blocks the quickening of
field accesses instead of blocking all optimizations. (Not sure this
is worth keeping around.)
Change-Id: I7f7c658e3b682c7251cdf17cae58d79bd04ba2a0
Elliott Hughes [Tue, 6 Apr 2010 01:13:52 +0000 (18:13 -0700)]
An InlineNative for String.isEmpty, so it's not slower than length() == 0.
Before:
benchmark ns logarithmic runtime
IsEmpty 115 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
LengthEqualsZero 21 XXXXX||||||||||||||
With C intrinsic:
IsEmpty 30 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
LengthEqualsZero 20 XXXXXXXXXXXXXXXXXXXX||||||
With assembler intrinsic:
IsEmpty 15 XXXXXXXXXXXXXXXXXXXX||||||
LengthEqualsZero 21 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
(All times on passion.)
Change-Id: Ifcc37fe7b8efdd377675a448e0085e490d6767bc
Elliott Hughes [Thu, 8 Apr 2010 20:58:06 +0000 (13:58 -0700)]
Merge "resolved conflicts for merge of
19252faf to dalvik-dev" into dalvik-dev
Elliott Hughes [Thu, 8 Apr 2010 20:57:31 +0000 (13:57 -0700)]
Merge "Show a summary of failures at the end of a vogar run." into dalvik-dev
Elliott Hughes [Thu, 8 Apr 2010 20:57:04 +0000 (13:57 -0700)]
resolved conflicts for merge of
19252faf to dalvik-dev
Change-Id: Ibf429f9d9d048c01cb8f33b15e8b404d83138b74
Brian Carlstrom [Thu, 8 Apr 2010 20:52:15 +0000 (13:52 -0700)]
Merge "Fixing javax.net.ssl test failures on RI" into dalvik-dev
Elliott Hughes [Thu, 8 Apr 2010 20:11:14 +0000 (13:11 -0700)]
Show a summary of failures at the end of a vogar run.
Super-useful when running a large batch of jtreg tests, and also useful
for comparing against an earlier run (though hopefully we'll have a better
solution for that eventually).
Change-Id: I144f72ea1ae5240393b0b33193d7773537b7ee20
Elliott Hughes [Thu, 8 Apr 2010 20:06:14 +0000 (13:06 -0700)]
Merge "Add Java 6's new java.util.zip API." into dalvik-dev
Brian Carlstrom [Thu, 8 Apr 2010 18:58:15 +0000 (11:58 -0700)]
Fixing javax.net.ssl test failures on RI
vogar wasn't working at checkin on the host so I couldn't easily
retest these changes that came from the code review. Sure enought
there were some problems.
SSLSocketTest now uses a condition variable to synchronize between
the HandshakeCompletedListener and the main thread since the RI
does the callback asynchronously.
Back to ussing Arrays.equals for comparing Session ID byte[] since
Object.equals does not work.
Change-Id: I667b60dc065438efe4e3728d7a44b446ebc15e64
Elliott Hughes [Thu, 8 Apr 2010 01:27:01 +0000 (18:27 -0700)]
Add Java 6's new java.util.zip API.
This passes all but four of the harmony tests. The others fail because our
Deflater works differently to theirs; we already fail the corresponding
Deflater test.
The only jtreg test for ZipError neither passes nor fails: it appears that
this can only be thrown on Windows.
Bug:
2497395
Change-Id: I79687495da42507dda692c890ec939bdbc9be2b3
Elliott Hughes [Thu, 8 Apr 2010 18:10:54 +0000 (11:10 -0700)]
Disable System.setSecurityManager.
We plan on removing SecurityManager checks, which are expensive and non-useful,
especially because Android doesn't use SecurityManager itself. As a first step,
let's ship a release where it's no longer possible to set a SecurityManager.
Bug:
1320501
Change-Id: I88664dbc1d8b087579d54003a1aaed7b3d8412be
Jesse Wilson [Thu, 8 Apr 2010 17:40:43 +0000 (10:40 -0700)]
am
214e2855: am
5dde0c4d: Merge "Adding support for hash \'#\' comments to our JSON parser." into froyo
Merge commit '
214e285536e18e762c3207fecf97dacdd22e472e' into dalvik-dev
* commit '
214e285536e18e762c3207fecf97dacdd22e472e':
Adding support for hash '#' comments to our JSON parser.
Andy McFadden [Thu, 8 Apr 2010 17:40:38 +0000 (10:40 -0700)]
am
abfb34bc: am
c7b7edf7: Merge "Trivial doc change: froyo is 2.2." into froyo
Merge commit '
abfb34bcda2e2320121850a9cb5b10e8f10e9833' into dalvik-dev
* commit '
abfb34bcda2e2320121850a9cb5b10e8f10e9833':
Trivial doc change: froyo is 2.2.
Andy McFadden [Tue, 6 Apr 2010 20:45:38 +0000 (13:45 -0700)]
Trivial doc change: froyo is 2.2.
Change-Id: Icce3770dd318ec8853d5120c13a85656ed1171ea
Jesse Wilson [Wed, 7 Apr 2010 18:38:29 +0000 (11:38 -0700)]
Merge "Fixing a vogar issue where outcomes weren't making their way into XML." into dalvik-dev
Jesse Wilson [Wed, 7 Apr 2010 18:37:56 +0000 (11:37 -0700)]
Merge "Bring the latest Caliper (r102) into vogar." into dalvik-dev
Jesse Wilson [Wed, 7 Apr 2010 17:53:21 +0000 (10:53 -0700)]
Bring the latest Caliper (r102) into vogar.
Jesse Wilson [Wed, 7 Apr 2010 17:52:28 +0000 (10:52 -0700)]
Fixing a vogar issue where outcomes weren't making their way into XML.
Andy McFadden [Wed, 7 Apr 2010 14:27:55 +0000 (07:27 -0700)]
Merge "Correct long-standing thread status change bug." into dalvik-dev
Andy McFadden [Wed, 7 Apr 2010 14:27:30 +0000 (07:27 -0700)]
Merge "Rearrange some things." into dalvik-dev
Jesse Wilson [Wed, 7 Apr 2010 03:25:23 +0000 (20:25 -0700)]
Merge "Adding support for command line args to vogar-initiated processes." into dalvik-dev
Elliott Hughes [Wed, 7 Apr 2010 03:17:36 +0000 (20:17 -0700)]
Fix build by adding missing imports.
Change-Id: I433778b02d3ad67090ee88444700f79b3138d778
Jesse Wilson [Wed, 7 Apr 2010 03:11:13 +0000 (20:11 -0700)]
Adding support for command line args to vogar-initiated processes.
Also adding monitor timeouts for use by the continuous build.
From the huge number of files in this relatively simple CL, it's
becoming clear that we need to simplify vogar's internals!