Community
Participate
Working Groups
RM TPTP Junit failing in test automation
I imagine this is a setup issue of some sort. We are able to run the JUnits in development and pass approx 98% of tests. Please confirm you have updated all properties files in your workspace and that all referenced directories exist on your machine. In some cases you may have to export a folder out of eclipse to your hard drive.
Status update, Saurabh?
After fixing some code in test automation I am getting following test status for RM junits. (1) org.eclipse.cosmos.rm.validation.tests : 97.2% pass, 2.8% fail. (2) org.eclipse.cosmos.rm.repository.tests : 57.4% pass, 31.2% fail, 11.3% error. I checked different properties file and they seems to look fine.
(In reply to comment #3) > After fixing some code in test automation I am getting following test status > for RM junits. > > (1) org.eclipse.cosmos.rm.validation.tests : 97.2% pass, 2.8% fail. > (2) org.eclipse.cosmos.rm.repository.tests : 57.4% pass, 31.2% fail, 11.3% > error. > > I checked different properties file and they seems to look fine. > The validation failure rate seems reasonable, or at least not too far off. However, there seems to be a config problem with the repo tests. John, do you have any ideas on what he might be lacking in his config, or do you need to see some specific test failure messages?
(In reply to comment #4) > John, do you > have any ideas on what he might be lacking in his config, or do you need to see > some specific test failure messages? I'm a little confused as to how anyone could be expected to do much given the data provided in this bug. I can provide the configuration I used along with the results. Comparing my workspace to CVS, on the repository.* and validation.* projects, I have the following local mods /org.eclipse.cosmos.rm.repository.tests/src/META-INF configs 2,4,5,6,8 - machine-specific dir paths /org.eclipse.cosmos.rm.validation.tests/src/org/eclipse/cosmos/rm/internal/validation/common/testMessages.properties - commented out test dir line so it uses the Java temp dir. This was not a local mod before, so I intuit that this was the "fix" to build automation mentioned before. The copyright supports this hypothesis. .classpath's (expected given my install situation) The rest(*) is exactly as in CVS according to Team Synchronizing. Having just re-run both AllTests again, I still get 100% successful on both. * namely org.eclipse.cosmos.common.tests org.eclipse.cosmos.rm.example.datacenter org.eclipse.cosmos.rm.example.rtx org.eclipse.cosmos.rm.smlif org.eclipse.cosmos.rm.smlif.tests They only intermittent errors I have seen in the past were schema validation failures where the error message said that a definition for xml:lang could not be found. In every such case, re-running the test again caused it to succeed. The root issue was that schema validation was fetching the schema definition for xml:lang over the network from W3C, and sometimes failing to retrieve it. Caching a local copy is the only reliable solution for that (theoretically it should be in the built-in classes, but it is not in some versions of Java, there is a Java bug open for that). Absent the xml:lang validation messages, I think we would need to understand where the reporter's workspace differs from CVS -and- see some representative diagnostic messages in order to progress any further.
