HBASE-30382: Update zstd-jni to 1.5.7-15 (branch-2.5) - #8656
arvindKandpal-ksolves wants to merge 1 commit into
Conversation
|
Verified locally on branch-2.5, JDK 17.0.20, Hadoop3 profile (-Dhadoop.profile=3.0), matching the "Yetus JDK17 Hadoop3 Unit Check / medium" failing job. With zstd-jni 1.5.7-15 (this PR): With zstd-jni 1.5.7-2 (pristine upstream/branch-2.5, pom.xml reverted): Same failure, identical assertion, with and without the zstd-jni bump — confirms this is a pre-existing issue on branch-2.5, unrelated to this PR. Also confirmed no reference to zstd/compression in either test file or the shared test base class. |
|
Verified locally on branch-2.5, JDK 11.0.32, Hadoop3 profile (-Dhadoop.profile=3.0), matching "Yetus JDK11 Hadoop3 Unit Check / medium". TestWALCompressionLzma passes. TestBufferedMutator#testMultiThread and TestBasicWALEntryStreamFSHLog#testCleanClosedWALs fail — confirmed pre-existing by reverting pom.xml to pristine zstd-jni 1.5.7-2 and reproducing the same failures. With zstd-jni 1.5.7-15: With zstd-jni 1.5.7-2 (pristine): |
|
Verified locally on branch-2.5, JDK 1.8.0_502, Hadoop2 (default profile), matching "Yetus JDK8 Hadoop2 Unit Check / medium". TestWALCompressionLzma, TestMetaCache, TestMasterRegionFlush, TestBindExceptionHandling: all pass. TestBasicWALEntryStreamAsyncFSWAL/FSHLog#testCleanClosedWALs: fail identically with and without the zstd-jni bump (reverted pom.xml to pristine 1.5.7-2 and reproduced) — pre-existing issue. TestBufferedMutator#testMultiThread: failed once with the bump, passed on rerun without it and again passed with try — consistent with CI's own "Flakes" classification (timing-sensitive, not deterministic either way). see here : |
JDK17 Hadoop3 (medium)All failed test passed JDK11 Hadoop3 (medium)this is failed test in this branch that is passed these marked as Flakes in failed report and again fail in local and it is not related to our changes. JDK8 Hadoop2 (medium)All failed test passed |
|
Closing. #8652 will be cherry-picked to this branch. |
What changes were proposed in this pull request?
Bumps zstd-jni from 1.5.7-2 to 1.5.7-15 in pom.xml (branch-2.5).
Why are the changes needed?
zstd-jni versions prior to 1.5.7-14 contain vulnerable code addressed by:
Does this PR introduce any user-facing change?
No.
How was this patch tested?
Ran the full zstd-jni compression test suite before and after the bump:
TestWALCompressionZstd, TestZstdCodec, TestHFileCompressionZstd,
TestZstdDictionary, TestZstdDictionarySplitMerge, TestZstdByteBuffDecompressor.
Baseline (1.5.7-2): 15 tests run, 0 failures, 0 errors
After bump (1.5.7-15): 15 tests run, 0 failures, 0 errors
Verified via
mvn dependency:treethat the resolved artifact iscom.github.luben:zstd-jni:jar:1.5.7-15.
Note: built with
-Dhadoop.profile=3.0(required for JDK11+).