we used InlineIVs with great success so far, we saw almost adoubling in speed.
However, a while ago a distribution zip created by my colleague failed an error in the bigdata.war. There was a problem serializing a string in the database using our custom InlineIVs-class.
Looking into the matter we found there was a diffference between his compiled version of our InlineIV-class and my version. It had to do with accented characters. Using WinMerge we saw this:
Looks like a ISO-8859-1 vs Unicode encoding issue. I have fixed our building process to specify the encoding and that seems to have resolved the issue. Apparently specifying the encoding is important in compiling a custom InlineIVs-class.
However, there was another problem we noticed. Different versions of Java generate different values for the InlineIV-strings. This seems to be caused by the place of the "class" entry for our custom class, in java major version 52 it is generated before all strings, in Java major version after all strings:
major version 52:
#8 = Class #282 // nl/overheid/linkeddata/blazegraph/LinkedDataVocabularyDecl
major version 51:
#243 = Class #517 // nl/overheid/linkeddata/blazegraph/LinkedDataVocabularyDecl
Due to the unpredictability we have removed the InlineIVs for now. While I have reported this as a bug, is there anything we can do to fix this, or de we have to wait for a new release with a fix?