Sergey Kuksenko wrote: > Hi, All! > Fix for " > org.apache.harmony.nio.tests.java.nio.channels.FileLockInterruptionExceptionTest" > > failure are avaliable in > https://issues.apache.org/jira/browse/HARMONY-4017 > > The problem was not related to URLClassLoader. URLClassLoader only opened > the problem. > First of all I should note thate FileLockInterruptionExceptionTest wasn't > failed in separate run. > This failure was only in run of all tests. > When we read from InputStream with "read" method the only way to understand > that we got EOF is getting "-1" as return value for "read". > "hyfile_read" set "the last hyerror value" when some IO falure occures. > But "hyfile_read" doesn't clean "the last hyerror value" after > succesfull IO > operations. > And return value of "hyfile_read" is "-1" both for failed IO operation and > getting OEF. > Some test with errors was run before FileLockInterruptionExceptionTest. > And then we tried to read FileLockInterruptionExceptionTest.class file. > read it. and got "-1" for OEF but this "-1" was treated as errors, because > "the last hyerror value" was not empty. > > And it is an example to "even number of bugs" theory. :)
Good work in tracking that one down Sergey! That has obviously been lurking for a while. Regards, Tim
