Sorting on Score Problem

[Kuai, Ben]

Hi

We met a wired problem in our project when sorting by score in Solr 4.0, the 
biggest score document is not a the top the debug explanation from solr are 
like this,

First Document
1.8412635 = (MATCH) sum of:
  2675.7964 = (MATCH) sum of:
    0.0 = (MATCH) sum of:
      0.0 = (MATCH) max of:
        0.0 = (MATCH) btq, product of:
          0.0 = weight(nameComplexNoTfNoIdf:plumber^0.0 in 0) [], result of:
            0.0 = score(doc=0,freq=1.0 = phraseFreq=1.0
            ..........

Second Document
1.8412637 = (MATCH) sum of:
  0.26757964 = (MATCH) sum of:
    0.0 = (MATCH) sum of:
      0.0 = (MATCH) max of:
        0.0 = (MATCH) btq, product of:
          0.0 = weight(nameComplexNoTfNoIdf:plumber^0.0 in 0) [], result of:
            0.0 = score(doc=0,freq=1.0 = phraseFreq=1.0
            .................

Third Document
1.841253 = (MATCH) sum of:
  2675.7964 = (MATCH) sum of:
    0.0 = (MATCH) sum of:
      0.0 = (MATCH) max of:
        0.0 = (MATCH) btq, product of:
          0.0 = weight(nameComplexNoTfNoIdf:plumber^0.0 in 0) [], result of:
            0.0 = score(doc=0,freq=1.0 = phraseFreq=1.0
            .......


Then we thought it could be a float rounding problem then we implement our own 
similarity class to increse queryNorm by 10,000 and it changes the score scale 
but the rank is still wrong.

Dose Anyone have the similiar issue?

I can debug with solr source code and please shed some light on the sorting part

Thanks