September 2021 - core-libs-dev

RFR: 8274464: Remove redundant stream() call before forEach in java.* modules
by Andrey Turbanov 06 Oct '21

06 Oct '21

8274464: Remove redundant stream() call before forEach in java.* modules ------------- Commit messages: - [PATCH] Remove redundant stream() call before forEach in java.* modules Changes: https://git.openjdk.java.net/jdk/pull/5520/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=5520&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8274464 Stats: 8 lines in 4 files changed: 0 ins; 3 del; 5 mod Patch: https://git.openjdk.java.net/jdk/pull/5520.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/5520/head:pull/5520 PR: https://git.openjdk.java.net/jdk/pull/5520

4 4

RFR: 8274397: [macOS] Stop setting env. var JAVA_MAIN_CLASS_<pid> in launcher code
by Phil Race 05 Oct '21

05 Oct '21

macOS launcher code sets JAVA_MAIN_CLASS_<pid> which is read by AWT to set the name of the application in the system menu bar. Because this set shortly after the VM is running, it causes a thread safety issue described in https://bugs.openjdk.java.net/browse/JDK-8270549 Since the AWT already looks for the system property "apple.awt.application.name" for this same purpose, we can set that instead of the env. var and avoid the threading issue. This trades the JAVA_MAIN_CLASS_<pid> pollution of the environment for setting a system property which is visible to apps as well but it seems a reasonable trade-off since we already (implicitly) support this variable anyway in preference to the env. var. ------------- Commit messages: - 8274397: [macOS] Stop setting env. var JAVA_MAIN_CLASS_<pid> in launcher code - 8274397: [macOS] Stop setting env. var JAVA_MAIN_CLASS_<pid> in launcher code Changes: https://git.openjdk.java.net/jdk/pull/5724/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=5724&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8274397 Stats: 75 lines in 2 files changed: 33 ins; 33 del; 9 mod Patch: https://git.openjdk.java.net/jdk/pull/5724.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/5724/head:pull/5724 PR: https://git.openjdk.java.net/jdk/pull/5724

4 15

RFR: 8274299: Make Method/Constructor/Field accessors @Stable
by Peter Levart 05 Oct '21

05 Oct '21

This patch improves reflective access speed as shown by the included benchmarks: https://jmh.morethan.io/?gists=902f4b43519c4f96c7abcd14cdc2d27d,ac490481e30… ... and is also a prerequisite to make JEP 416 (Reimplement Core Reflection with Method Handle) perform better in some circumstances. ------------- Commit messages: - Add JMH benchmarks - Make Method/Constructor/Field accessors @Stable Changes: https://git.openjdk.java.net/jdk/pull/5694/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=5694&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8274299 Stats: 721 lines in 13 files changed: 637 ins; 37 del; 47 mod Patch: https://git.openjdk.java.net/jdk/pull/5694.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/5694/head:pull/5694 PR: https://git.openjdk.java.net/jdk/pull/5694

4 27

RFR: 8274079: Cleanup unnecessary calls to Throwable.initCause() in java.base module
by Andrey Turbanov 05 Oct '21

05 Oct '21

Pass "cause" exception as constructor parameter is shorter and easier to read. ------------- Commit messages: - [PATCH] Cleanup unnecessary calls to Throwable.initCause() in java.base module Changes: https://git.openjdk.java.net/jdk/pull/5551/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=5551&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8274079 Stats: 133 lines in 22 files changed: 0 ins; 77 del; 56 mod Patch: https://git.openjdk.java.net/jdk/pull/5551.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/5551/head:pull/5551 PR: https://git.openjdk.java.net/jdk/pull/5551

3 3

RFR: 8274606: Fix jaxp/javax/xml/jaxp/unittest/transform/SurrogateTest.java test
by Alex Kasko 04 Oct '21

04 Oct '21

I was working on backporting JDK-8268457 and found minor problems with the test introduced there: 1. `compareWith*` helper methods are used without `Assert.assertTrue()` wrapping, so they are effectively ignored 2. `this.getClass().getResourceAsStream()` is used to load test input data, it actually returns null in test run, so transformation is done without input data Note, that SurrogateTest.zip reproducer attached to JDK-8268457 is valid and fully functional, problems likely were introduced when it was adapted into test. The change is to the test only, it wraps `compareWith*` helpers and loads data the same way as XSL is loaded. Additionally `compareStringWithGold` was changed to `compareLinesWithGold` to exclude possible line endings problems. Testing: checked that the test fails when JDK-8268457 code fix is reverted, checked that is passes on master. ------------- Commit messages: - 8274606: Fix jaxp/javax/xml/jaxp/unittest/transform/SurrogateTest.java test Changes: https://git.openjdk.java.net/jdk/pull/5779/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=5779&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8274606 Stats: 10 lines in 1 file changed: 3 ins; 1 del; 6 mod Patch: https://git.openjdk.java.net/jdk/pull/5779.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/5779/head:pull/5779 PR: https://git.openjdk.java.net/jdk/pull/5779

3 4

Discussion: easier Stream closing
by Tagir Valeev 03 Oct '21

03 Oct '21

Hello! With current NIO file API, a very simple problem to get the list of all files in the directory requires some ceremony: List<Path> paths; try (Stream<Path> stream = Files.list(Path.of("/etc"))) { paths = stream.toList(); } If we skip try-with-resources, we may experience OS file handles leak, so it's desired to keep it. Yet, it requires doing boring stuff. In this sense, classic new File("/etc").list() was somewhat more convenient (despite its awful error handling). I like how this problem is solved in StackWalker API [1]: it limits the lifetime of the Stream by requiring a user-specified function to consume it. After the function is applied, the stream is closed automatically. We could add a similar overload to the Files API. As an additional feature, we could also translate all UncheckedIOException's back to IOException for more uniform exception processing: /** * @param dir The path to the directory * @param function function to apply to the stream of directory files * @param <T> type of the result * @return result of the function * @throws IOException if an I/O error occurs when opening the directory, or * UncheckedIOException is thrown by the supplied function. */ public static <T> T list(Path dir, Function<? super Stream<Path>, ? extends T> function) throws IOException { try (Stream<Path> stream = Files.list(dir)) { return function.apply(stream); } catch (UncheckedIOException exception) { throw exception.getCause(); } } In this case, getting the List of all files in the directory will be as simple as List<Path> paths = Files.list(Path.of("/etc"), Stream::toList); This doesn't limit the flexibility. For example, if we need only file names instead of full paths, we can do this: List<Path> paths = Files.list(Path.of("/etc"), s -> s.map(Path::getFileName).toList()); Alternatively, we could enhance the BaseStream interface in a similar way. It won't allow us to translate exceptions, but could be useful for every stream that must be closed after consumption: // in BaseStream: /** * Apply a given function to this stream, then close the stream. * No further operation on the stream will be possible after that. * * @param function function to apply * @param <R> type of the function result * @return result of the function * @see #close() */ default <R> R consumeAndClose(Function<? super S, ? extends R> function) { try(@SuppressWarnings("unchecked") S s = (S) this) { return function.apply(s); } } The usage would be a little bit longer but still more pleasant than explicit try-with-resources: List<Path> list = Files.list(Path.of("/etc")).consumeAndClose(Stream::toList); On the other hand, in this case, we are free to put intermediate operations outside of consumeAndClose, so we won't need to nest everything inside the function. Only terminal operation should be placed inside the consumeAndClose. E.g., if we need file names only, like above, we can do this: List<Path> list = Files.list(Path.of("/etc")).map(Path::getFileName).consumeAndClose(Stream::toList); What do you think? [1] https://docs.oracle.com/en/java/javase/17/docs/api/java.base/java/lang/Stac…

4 7

RFR: 4890732: GZIPOutputStream doesn't support, in fact thwarts, use of optional GZIP fields
by Lin Zang 30 Sep '21

30 Sep '21

4890732: GZIPOutputStream doesn't support, in fact thwarts, use of optional GZIP fields ------------- Commit messages: - remove trailing spaces - 4890732: support optional GZIP fields in GZIPOutputStream Changes: https://git.openjdk.java.net/jdk/pull/3072/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=3072&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-4890732 Stats: 184 lines in 1 file changed: 184 ins; 0 del; 0 mod Patch: https://git.openjdk.java.net/jdk/pull/3072.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/3072/head:pull/3072 PR: https://git.openjdk.java.net/jdk/pull/3072

5 46

RFR: 8274506: TestPids.java and TestPidsLimit.java fail with podman run as root
by Severin Gehwolf 30 Sep '21

30 Sep '21

Please review this test fix to work around a podman issue[1]. `podman` expects for the "unlimited" option of `--pids-limit` to be `0` whereas `docker` wants `-1`. See the JBS bug for details. Thoughts? Testing: hotspot/jdk container tests with docker and podman. Two pids tests used to fail and pass with the patch. [1] https://github.com/containers/podman/issues/11782 ------------- Commit messages: - 8274506: TestPids.java and TestPidsLimit.java fail with podman run as root Changes: https://git.openjdk.java.net/jdk/pull/5750/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=5750&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8274506 Stats: 12 lines in 2 files changed: 10 ins; 0 del; 2 mod Patch: https://git.openjdk.java.net/jdk/pull/5750.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/5750/head:pull/5750 PR: https://git.openjdk.java.net/jdk/pull/5750

3 4

RFR: 8274523: java/lang/management/MemoryMXBean/MemoryTest.java test should handle Shenandoah
by Aleksey Shipilev 30 Sep '21

30 Sep '21

Currently it fails with: $ CONF=linux-x86_64-server-fastdebug make run-test TEST=java/lang/management/MemoryMXBean/MemoryTest.java STDERR: java.lang.RuntimeException: TEST FAILED: Number of heap pools = 1 but expected <= 3 and >= 3 Z already handles it with a special configuration, Shenandoah should do the same. Additional testing: - [x] Affected test now works for Shenandoah - [x] Affected test still works for Z - [x] Affected test still works for G1 ------------- Commit messages: - Fix Changes: https://git.openjdk.java.net/jdk/pull/5758/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=5758&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8274523 Stats: 2 lines in 1 file changed: 0 ins; 0 del; 2 mod Patch: https://git.openjdk.java.net/jdk/pull/5758.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/5758/head:pull/5758 PR: https://git.openjdk.java.net/jdk/pull/5758

3 4

RFR: 8273435: Remove redundant zero-length check in ClassDesc.of
by Andrey Turbanov 29 Sep '21

29 Sep '21

After [JDK-8215510](https://bugs.openjdk.java.net/browse/JDK-8215510) (eed3a536c0) this condition is always `false`. Empty package name is handled separately. Found by IntelliJ inspection. ------------- Commit messages: - [PATCH] Remove redundant zero-length check in ClassDesc.of Changes: https://git.openjdk.java.net/jdk/pull/5157/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=5157&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8273435 Stats: 2 lines in 1 file changed: 0 ins; 0 del; 2 mod Patch: https://git.openjdk.java.net/jdk/pull/5157.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/5157/head:pull/5157 PR: https://git.openjdk.java.net/jdk/pull/5157

3 4