Clang static analyzer results / wine-1.5.28-66-g6899279
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(a)17976891177308-0d34-0410-b5e6... clang version: git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk(a)17977191177308-0d34-0410-b5e6-... ./configure --disable-tests was used to reduce noise [austin(a)localhost ~]$ sha1sum scan-build-2013-04-18-2.tar.bz2 51e28c96a04ebe9ef937678149f9e7e3ed4cb20f scan-build-2013-04-18-2.tar.bz2 [austin(a)localhost ~]$ du -h scan-build-2013-04-18-2.tar.bz2 35M scan-build-2013-04-18-2.tar.bz2 http://www.mediafire.com/?xwb05d6er0aclfe -- -Austin
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(a)179768 91177308-0d34-0410-b5e6-96231b3b80d8 clang version: git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk(a)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
On Fri, Apr 19, 2013 at 1:52 AM, Michael Stefaniuc <mstefani(a)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(a)179768 91177308-0d34-0410-b5e6-96231b3b80d8 clang version: git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk(a)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 -- -Austin
On Wed, Apr 24, 2013 at 7:40 AM, Austin English <austinenglish(a)gmail.com>wrote:
On Fri, Apr 19, 2013 at 1:52 AM, Michael Stefaniuc <mstefani(a)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(a)179768 91177308-0d34-0410-b5e6-96231b3b80d8 clang version: git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk(a)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. 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. Frédéric
On 04/24/13 11:35, Frédéric Delanoy wrote:
On Wed, Apr 24, 2013 at 7:40 AM, Austin English <austinenglish(a)gmail.com <mailto:austinenglish(a)gmail.com>> wrote:
On Fri, Apr 19, 2013 at 1:52 AM, Michael Stefaniuc <mstefani(a)redhat.com <mailto:mstefani(a)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(a)179768 >> 91177308-0d34-0410-b5e6-96231b3b80d8 >> clang version: git-svn-id: >> https://llvm.org/svn/llvm-project/cfe/trunk(a)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
On Apr 24, 2013, at 5:10 AM, Jacek Caban wrote:
On 04/24/13 11:35, Frédéric Delanoy wrote:
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.
Would it work to do two runs and have two different result sets? I think that "make install" rather than "make" avoids building the tests without requiring --disable-tests. (You can set DESTDIR to some temp directory in your home or /tmp.) After that, a plain "make" will make just the tests. -Ken
On Wed, Apr 24, 2013 at 3:28 AM, Ken Thomases <ken(a)codeweavers.com> wrote:
On Apr 24, 2013, at 5:10 AM, Jacek Caban wrote:
On 04/24/13 11:35, Frédéric Delanoy wrote:
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.
Would it work to do two runs and have two different result sets?
I think that "make install" rather than "make" avoids building the tests without requiring --disable-tests. (You can set DESTDIR to some temp directory in your home or /tmp.) After that, a plain "make" will make just the tests.
-Ken
If people find results with the tests enabled useful, I can certainly run it twice, yes. -- -Austin
On Wed, Apr 24, 2013 at 10:39 AM, Austin English <austinenglish(a)gmail.com> wrote:
On Wed, Apr 24, 2013 at 3:28 AM, Ken Thomases <ken(a)codeweavers.com> wrote:
On Apr 24, 2013, at 5:10 AM, Jacek Caban wrote:
On 04/24/13 11:35, Frédéric Delanoy wrote:
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.
Would it work to do two runs and have two different result sets?
I think that "make install" rather than "make" avoids building the tests without requiring --disable-tests. (You can set DESTDIR to some temp directory in your home or /tmp.) After that, a plain "make" will make just the tests.
-Ken
If people find results with the tests enabled useful, I can certainly run it twice, yes.
-- -Austin
[austin(a)localhost ~]$ sha1sum scan-build-2013-04-24-with-tests.tar.bz2 9dfd540f2ac2bd656ab326bc9c5fff7240f966b6 scan-build-2013-04-24-with-tests.tar.bz2 [austin(a)localhost ~]$ du -h scan-build-2013-04-24-with-tests.tar.bz2 54M scan-build-2013-04-24-with-tests.tar.bz2 http://www.mediafire.com/?jtiqxdq8fql2b9l These results are also against wine-1.5.28-66-g6899279. -- -Austin
participants (5)
-
Austin English -
Frédéric Delanoy -
Jacek Caban -
Ken Thomases -
Michael Stefaniuc