On 04/24/13 11:35, Frédéric Delanoy wrote:
On Wed, Apr 24, 2013 at 7:40 AM, Austin English <austinenglish@gmail.com> wrote:
On Fri, Apr 19, 2013 at 1:52 AM, Michael Stefaniuc <mstefani@redhat.com> wrote:
>
> On 04/19/2013 12:43 AM, Austin English wrote:
>>
>> With wine-1.5.20 and clang 3.2, the test suite is in the same state on
>> my Fedora 18 machine as gcc-4.7.2
>>
>> llvm version:     git-svn-id:
>> https://llvm.org/svn/llvm-project/llvm/trunk@179768
>> 91177308-0d34-0410-b5e6-96231b3b80d8
>> clang version: git-svn-id:
>> https://llvm.org/svn/llvm-project/cfe/trunk@179771
>> 91177308-0d34-0410-b5e6-96231b3b80d8
>>
>> ./configure --disable-tests was used to reduce noise
>
> Static analyse of the tests is important as well. Over the years I found bad tests that didn't test what they were supposed to test. And even cases where the Wine code was bad due to that as it "passed" the tests.
>
> bye
>         michael


That was suggested by Jacek, see
http://www.winehq.org/pipermail/wine-devel/2012-January/094059.html

Sure, but Michael has a good point here IMHO.

Yeah, if it finds a real problem, then I agree. The problem with reports from tests is that there are tons of things like error handling that we don't intend to 'fix', so they do (and always will) make it harder to find similar real bugs in the rest of code base. Plus we're never going to be anywhere close to 0 problems this way ;)

Although, one can sort/filter by file, or if it's really that annoying, a quick script removing <tr> elements containing "/tests" should be easy enough.

Well... that's far from perfect, but could work.

Cheers,
Jacek