Uploaded image for project: 'Lucene - Core'
  1. Lucene - Core
  2. LUCENE-5537

Fix version check in TestCheckIndexes



    • Bug
    • Status: Closed
    • Blocker
    • Resolution: Fixed
    • 4.7
    • 4.7.1, 4.8, 6.0
    • core/other
    • None
    • New


      The crazy version check in TestCheckIndexes is broken. Whenever it fails, it requires a horrible amount of understanding:

      • error messages are useless
      • you don't know what it really checks and why (no documentation)

      In addition, we have no clear workflow how to handle version updates:

      • In 4.6 series we never set the dev.version in common-build to 4.6.1, it always stayed 4.6. The test worked therefore, because the LUCENE_MAIN_VERSION (which is always "x.y") was identical
      • In 4.7 steve_rowe changed the version to the real release version in the branch. The test failed because of this change (LUCENE_MAIN_VERSION was not identical as the dotted bugfix version)

      We should in any case fix the test:

      • move it out of TestCheckIndex, it should be in oal.util.TestConstants
      • Be more verbose on loggin
      • Remove special cases (leftovers from 4.0-BETA) where we did crazy stuff to make this test pass in alphas and betas.
      • Only add a check to LUCENE_MAIN_VERSION that it is in format "x.y" (which is required by Lucene index format), and only check that it is the prefix of the real version as reported by common_build.xml.

      We should also document and write down when we change the version numbers. I would prefer to use the variant of Lucene 4.6: never change the version number in common-build and "override" it while building artifacts.


        1. LUCENE-5537.patch
          8 kB
          Uwe Schindler
        2. LUCENE-5537.patch
          7 kB
          Uwe Schindler
        3. LUCENE-5537.patch
          7 kB
          Uwe Schindler
        4. LUCENE-5537.patch
          6 kB
          Uwe Schindler

        Issue Links



              uschindler Uwe Schindler
              uschindler Uwe Schindler
              0 Vote for this issue
              2 Start watching this issue