I tried to investigated the issue but have not succeed. (1) Since it is an intermittent error in my machines, I guess that it is a threading-related issue. I then tried previous of drlvm and could still observe the intermittent failure in M5 (r629320). I also checked RI and VME and have not seen such problem. (2) Observed same stacktrace, I see there is no problem to parse such files in a single-threaded environment, same as what Charles observed. Looking more in EUT35 source, I see that EUT open the file with BufferedInputStream and then pass the stream to SAXParser. The stacktrace itself likely shows that we can read nothing from the BufferedInputStream. I guessed that there might be some problems in drlvm when cooperated with the threaded BufferedInputStream but have not verified it.
On Thu, Feb 26, 2009 at 1:11 PM, Charles Lee <[email protected]> wrote: > Hi guys, > I'm running EUT releng test on the linux machine. The test is always > failed. Here is something I found, and need some help: > 1. I have encountered a " [Fatal Error] :1:20: The version is required in > the XML declaration." when running java-test target just like what Lai > ChunRong mentioned in > HARMONY-6062<http://issues.apache.org/jira/browse/HARMONY-6062>, > but without any luck to pass the test occasionally. > 2. With the -consolelog, I found the log file (attached), and says > "java.io.IOException: The version is required in the XML declaration." The > exception is thrown by the eclipse ProfileParser, which is parsing a profile > (attached). The profile is complete. The stack trace contains no information > about harmony. With these information, I guess there maybe something wrong > when the SAXParser parse a file. So I have run a test to check SAXParser > only to find there is no problem to parse the same log. > 3. I have tested the relEng with RI. RI passed the configuration phase, but > will setup a eclipse workbench and never close it. Harmony with current ibm > vm behaves the same. > 4. I have tested the relEng on windows. The relEng test passed but behavior > different from what it does on linux. No workbench is setup. I have checked > the test.xml about the relEng test, find this line in the setupcompare: > "only run compare test on Linux machine since it is the fastest > machine. Also, the compare tool only parses feature.xmls so it only needs to > be run on one machine" > On linux, target setupcompare will be run. It will download a eclipse > with older version and change some configuration to compare with the newest > one. I guess here is reason what makes difference. > > Any idea? > > -- > Yours sincerely, > Charles Lee > > > > > -- > Yours sincerely, > Charles Lee > >
