Re: Intermittent RequestQueue problem

From: Andrus (andru..bjectstyle.org)
Date: Wed Dec 04 2002 - 21:25:39 EST

  • Next message: Andrus: "Alpha 5 Release planning"

    Craig,

    this was already reported by Holger, (with some of my unproven ideas in
    comments :-)) :

    http://sourceforge.net/tracker/index.php?func=detail&aid=646399&group_id=48132&atid=452068

    (I wonder if there is a way to configure sourceforge trackers to send all
    bug report emails to cayenne-cvs?)

    Anyway, RequestQueue itself is database independent and does not perform
    any database access (so this narrows the possibilities). Also I believe
    that the problem is somewhere in the test. It was a rather brain damaging
    exercise to write unit tests for the multithreaded code to begin with :-).
    So there may be some synchronization timing issues that I overlooked.

    Andrus

    At 03:11 PM 12/5/2002 +1300, Craig Miskell wrote:
    >Hi,
    > Just thought I'd report this, in case it rings bells with anyone
    > else.
    >
    >I came into work this morning, and got the output from the automated
    >unittest run from last night (runs at midnight, on my dev workstation).
    >I see there was no updates from cvs, give a cursory glance to output of
    >tests and discover that RequestQueueTst failed at line 135 (asserting
    >that the results of testCase2 (second thread) is not null). It failed
    >for both Oracle 8.0.3 and 8i (two runs, right after each other). So i
    >run the tests manually, thinking it might be a timing issue, or
    >something odd happening around midnight. Same problem. I try against
    >my current dev checkout in case my other version is stuffed... same
    >issue.
    >
    >So I try a few more times (ant clean etc in between), still no luck.
    >Decide to do something else.
    >
    >Came back just now, and ran the tests again (for giggles). Still no
    >updates from CVS, but Lo! it runs fine, all tests pass.
    >
    >I have absolutely no idea what it was, but there *may* be a problem
    >somewhere in the RequestQueue code. Maybe. Possibly. A long shot :-)
    >Might depend on load, or some other odd thing that I can't quite put my
    >finger on.
    >
    >Craig



    This archive was generated by hypermail 2.0.0 : Wed Dec 04 2002 - 21:25:05 EST