Re: Updated high-unicode patch

On Thu, 30 Jun 2005, Oliver Jowett wrote:

> The attached patch is an updated and bugfixed version of my earlier
> patch to support UTF-8 encoding of unicode codepoints above U+FFFF

I'm don't like the cost of the new tests included here.  On my 2 x PIII
800 machine the entire JDBC 2 test suite takes 212 seconds.  testEncoding
takes 67 seconds testUTF8Decode takes 111 seconds for a whopping 84% of
the test time for 2 out of 220 tests.  Do you have any recommendations for
lowering this?  Random ideas that I could see:

  - using STEP in testUTF8Decode as well to reduce the number of iterations
    and adding a BIGSTEP for the high unicode range
  - selecting a representative sample of codepoints to test
  - creating a separate encoding test suite that is driven by "ant enctest"
  - running the tests only occasionally since they don't depend on JVM or
    database version and should therefore be pretty static, say randomly
    1 out of 10 runs.
  - doing nothing 3.5 minutes isn't really all that long

Kris Jurka