[Mono-devel-list] Corlib test cases

[David Sheldon]

The bugs remaining in the corlib test suite (apart from one or two
culture info changes that will be fixed next week) appear to fall into 4
main sections:

 1. The off-by-one error in the line numbers for errors (and the fact
    that the column is always reported as zero) - bug 45730

 2. Differences in the accuracy of our decimal math routines (I will
    include the lack of bankers rounding here).

 3. Tests for file-system semantics that are not provided by Unix. 
    
     -  Setting/getting the file creation time is not supported by 
        common unix file systems.

     -  We test that you get an exception if you try to open/get info
        on an already open file. It seems a shame to try and duplicate
        this failing of windows.

 4. The 14 new test failures for UTF-8 support that were committed
    yesterday.


Suites of test cases are significantly better if they are expected to
pass, allowing people to see when they have introduced more bugs.

I think that the tests for creation times/opening exceptions should be
restricted to mono on Windows. That should reduce the number of
failures. 

I am not sure what to do about the differences in our decimal
implementations. It is causing some tests on formatting numbers to fail,
due to the accuracy of the numbers, rather than failures in the number
printing routines (or maybe it is bugs in the routines introducing more
divides that necessary). Anyone got any suggestions here?

Do we have any idea when bug 45730 will be fixed? It does appear
non-trivial.

I'm assuming that the UTF-8 failure will be fixed in the near future. 

David
-- 
Give a man food, and he can eat for a day. Give a man a job, and he can only
eat for 30 minutes on break.
   -- Random sig file