Re: Additional normalization test cases

From: Michael \(michka\) Kaplan (
Date: Sat Mar 26 2005 - 17:53:19 CST

  • Next message: Jukka K. Korpela: "Symbols for chemical bonds"

    The normalization test cases are not intended to find random logic bugs in
    implementations; they are designed to find bugs that would make a random
    implementation non-conformant. They don't test for a NULL input param
    either, for example. But specifically they do not take every single "C" form
    to make sure they deocompse and every single "D" form to make sure they

    The D/KD in this case is:

        U+0041 U+0328 U+004c U+030c

    (tested with the .NET 2.0 implementation!)

    What was the bug you fixed? It might be easier to figure out a good way to
    add test cases by knowing what the implementation was doing....


    ----- Original Message -----
    From: "Elliotte Rusty Harold" <>
    To: <>
    Sent: Saturday, March 26, 2005 2:50 PM
    Subject: Additional normalization test cases

    > Is anyone collecting additional normalization test cases? This morning,
    > the normalizer I've been working on finally passed all 17,122 (times 5)
    > test cases in the Unicode suite, and then promptly died on the first
    > real data I threw at it. After some work, I was able to reduce the
    > trigger for my down to this two character string:
    > ĄĽ
    > That's
    > U+0104 U+013D
    > Apparently none of the 17,000 test cases managed to exercise the
    > particular bug in my code that made it fail on this two character
    > sequence (and many other strings as well). Any chance of adding this to
    > the test suite?
    > --
    > Elliotte Rusty Harold
    > XML in a Nutshell 3rd Edition Just Published!

    This archive was generated by hypermail 2.1.5 : Sat Mar 26 2005 - 17:53:02 CST