Community
Participate
Working Groups
Using: Eclipse 3.4 RC3, Mylyn 2.3.2, JDK6 Jira 3.12.3 I'm synchronizing with our Jira Server. That server is running with "crowd" for User management (might be relevant). I'm in fact retrieving all our items (several thousand). Some of them were created quite some time ago by people who meanwhile left. Now, while Jira is able to handle an entry which points to a reporter / assignee that does not exist any longer (in crowd), the Mylyn Jira integration obviously cannot (see attached screenshot for error message). Jira reports such an entry as either being reported from "anonymous" or remembers at least the username of the person but no details (no email etc.) I'd be glad, if Mylyn would behave similarly. Currently, the syncronisation completely aborts, thus I'm missing several thousand entries that I appear not to be able to retrieve somehow...
Created attachment 103938 [details] Error Message shown when Synchronizing with Jira
That looks an error from the SOAP service. I'll try to reproduce this.
Have you been able to reproduce the problem? Do you need some more informations?
Sorry Kay, I didn't get to this bug, yet. I'll schedule it for this week.
I can not reproduce this with our JIRA test repository that does not have crowd integration but I found these bugs in Atlassian's bug tracker: http://jira.atlassian.com/browse/CWD-202 http://jira.atlassian.com/browse/JRA-2073 The bug reports recommend to work around the problem by disabling users rather than deleting them in crowd. Since the error is caused on the server side it can unfortunately not be fixed in Mylyn. I have filed a bug against Atlassian here: http://jira.atlassian.com/browse/JRA-15399 . Kay, as a workaround you might be able to patch the RPC plugin yourself to supress the exception: http://confluence.atlassian.com/display/JIRAEXT/JIRA+RPC+plugin
I understand you cannot reproduce and it can be the problem of Atlassian, but don't you want at least, that Mylyn continues to retrieve the rest of the issues instead of aborting everything? Currently, this is the case: After the first Jira Issue causing such an exception, it just stops retrieving more issues from Jira, which is a pitty.
That's a good point. I'll take a look at the synchronization code.
Please note that the bug filed against JIRA (JRA-15399) has been fixed and will be included in v3.13 http://jira.atlassian.com/browse/JRA-15399 (In reply to comment #5) ... > > Since the error is caused on the server side it can unfortunately not be fixed > in Mylyn. I have filed a bug against Atlassian here: > http://jira.atlassian.com/browse/JRA-15399 . > ...
Thanks Mark for handling this so quickly! In addition I'll investigate if synchronization failures for a single task can be handled better on the Mylyn side to not affect synchronization of all tasks.
It seems to be the synchronize does remember tasks it failed on (done in the QueryHitCollector). It seems there is no exception thrown that would abort the parsing by Mylyn code. The synchonization job's "synchronizeQuery" method then checks the resuls of the whole query and only seems to remove removedChildren and set the timestamp. We might want to distinguish between a whole failed query and a query with some failed issues. In case of an issue making trouble the query could still be considered done, with a warning, but do everything else (remove children, set timestamp) But that does not even seem to be the problem here. I think the current code should finish the query and get all tasks right (except the ones giving trouble). But i might miss something here.
Thanks for looking into this Thomas. We have discussed API changes to support that better on bug 268467.
*** Bug 264929 has been marked as a duplicate of this bug. ***
I'll track the remainder of the Mylyn work left to improve handling of this error on bug 251658. Marking as NOT_ECLIPSE since this was fixed in JIRA 3.13 (comment 8).
