[
https://issues.apache.org/jira/browse/DERBY-3248?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12548391
]
Daniel John Debrunner commented on DERBY-3248:
----------------------------------------------
I wonder if this test can just use the NetworkServerTestSetup decorator to
start the server, it provides the option to start the server using a separate
process. No need to duplicate the code.
> SecureServerTest needs cleanup runServerCommand, possibly causing test
> failures when tests are run through ant.
> ---------------------------------------------------------------------------------------------------------------
>
> Key: DERBY-3248
> URL: https://issues.apache.org/jira/browse/DERBY-3248
> Project: Derby
> Issue Type: Bug
> Components: Test
> Affects Versions: 10.4.0.0
> Reporter: Daniel John Debrunner
>
> SecureServerTest.runServerCommand runs a command and then immediately creates
> a String based upon the output stream of the command. This string is then
> checked to see if it contains exception strings. However, I think that at the
> point the string is created there is no guarantee that the command has
> finished (or even started) execution. Thus there is a significant chance that
> the string will be empty (or at least missing exceptions) and thus always
> pass the test. Should the test be waiting for the command to complete using
> Process.waitFor().
> In getProcessOutput the bytes are converted to a String using the default
> encoding, this tends to cause problems for platforms with different encodings
> in test runs. It may be ok here since the output is from the system, if so it
> would be useful to have a comment indicating that the default encoding is ok
> here. (to avoid other tests blindly copying potentially error prone (due to
> encoding issues)code).
> In runServerCommand() if the exec fails the exception is just printed to
> System.err. SHould this instead be a fail() assertion. The test will fail if
> this occurs due to a NPE on the next access of serverProcess, but I think an
> explicit fail() would be more readable.
> I'm seeing 9 test failures when running derbynet._Suite through ant, on
> windows which I don't see if I run it directly using the swing test runner.
> This is one of the two tests that potentially cause issues (ie. removing it
> from the suite makes the failures disappear).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
