-
Bug
-
Resolution: Unresolved
-
P3
-
repo-loom
-
None
This test fails intermittently in the loom repo, usually loom-tier5 where the test runs with with fastdebug builds, -Xcomp, and JTREG_TEST_THREAD_FACTORY=Virtual (so the main thread is a virtual thread). The failure is usually linux-aarch64 but several sightings on macos-aarch64 and windows-x64.
When it fails then it's usually 3 out of the 8 tests fail with an exception like this:
Starting subtest 2.3: checkOutput TEXT allowOverlaps:false expectFound:true expectOrdered:false
FAILED: api/mC/p5/C5.html: output not as expected
at javadoc.tester.JavadocTester$OutputChecker.check(JavadocTester.java:1327)
at javadoc.tester.JavadocTester$OutputChecker.check(JavadocTester.java:1270)
at javadoc.tester.JavadocTester.checkOutput(JavadocTester.java:736)
at TestRedirectLinks.testWithServers(TestRedirectLinks.java:260)
at javadoc.tester.JavadocTester.runTest(JavadocTester.java:451)
It's possible that this testing mode isn't done with main line so it's possible this issue might duplicate with main line too. Nizar suggested it may be HTTP server related. Another possibility that there is a "short read", meaning less bytes are read from a socket than expected, could we be missing a check for the return from InputStream.read somewhere?
When it fails then it's usually 3 out of the 8 tests fail with an exception like this:
Starting subtest 2.3: checkOutput TEXT allowOverlaps:false expectFound:true expectOrdered:false
FAILED: api/mC/p5/C5.html: output not as expected
at javadoc.tester.JavadocTester$OutputChecker.check(JavadocTester.java:1327)
at javadoc.tester.JavadocTester$OutputChecker.check(JavadocTester.java:1270)
at javadoc.tester.JavadocTester.checkOutput(JavadocTester.java:736)
at TestRedirectLinks.testWithServers(TestRedirectLinks.java:260)
at javadoc.tester.JavadocTester.runTest(JavadocTester.java:451)
It's possible that this testing mode isn't done with main line so it's possible this issue might duplicate with main line too. Nizar suggested it may be HTTP server related. Another possibility that there is a "short read", meaning less bytes are read from a socket than expected, could we be missing a check for the return from InputStream.read somewhere?
- relates to
-
JDK-8338554 Fix inconsistencies in javadoc/doclet/testLinkOption/TestRedirectLinks.java
- Open