New valgrind results for Oct 21, 2009

[Dan Kegel]

I added suppressions for a few frequent spurious warnings and reran,
the updated results are in
 http://kegel.com/wine/valgrind/logs/2009-10-21-19.42.log
and broken out by test in
 http://kegel.com/wine/valgrind/logs/2009-10-21-19.42/
Please send me the suppression record for any warnings that are
spurious or benign, and mention the reason for each.

There's also an overall diff from the previous run,
http://kegel.com/wine/valgrind/logs/2009-10-21-19.42-diff.txt
but reading it is kind of a slog.

A few suggestions for people who would like to help fix these warnings:

1) Don't fix any error you don't fully understand.   Many
of these warnings are easy, but some are subtle.  When
in doubt, move on to an easier one.

2) Fixing leaks in test code is usually pretty safe, and
it's easier to get Alexandre to accept simple patches to tests,
so start with those if you don't have many patches under your belt.

3) the file
  http://kegel.com/wine/valgrind/logs/2009-10-21-19.42-count-by-file.txt
lists how many warnings are in each test; pick one with
just one error, and fix it.  Voila, one less failing test!

4) the file
 http://kegel.com/wine/valgrind/logs/2009-10-21-19.42-count-by-error.txt
lists how many of each warning there are.  Pick a warning
that happens a lot, and fix it.  Voila, lots fewer warnings!

5) if you know a warning is spurious, send it to me along with
the mangled version below it (that's in suppression format)
and why it's spurious or benign or not our fault, and I'll add it to
http://code.google.com/p/winezeug/source/browse/trunk/valgrind/valgrind-supp...

Have fun, be careful, and thanks to everyone who's pitching in!
- Dan

[Ben Klein]

2009/10/22 Dan Kegel <dank(a)kegel.com>:
> 2) Fixing leaks in test code is usually pretty safe, and
> it's easier to get Alexandre to accept simple patches to tests,
> so start with those if you don't have many patches under your belt.

I thought AJ didn't like leak fixes in the tests.

[Michael Stefaniuc]

Ben Klein wrote:
> 2009/10/22 Dan Kegel <dank(a)kegel.com>:
> > 2) Fixing leaks in test code is usually pretty safe, and
> > it's easier to get Alexandre to accept simple patches to tests,
> > so start with those if you don't have many patches under your belt.
>
> I thought AJ didn't like leak fixes in the tests.
> He doesn't like free's before an exit() or in short lived tools. But
> leak fixes in tests are accepted.

bye
	michael

[Nikolay Sivov]

Dan Kegel wrote:
> I added suppressions for a few frequent spurious warnings and reran,
> the updated results are in
>  http://kegel.com/wine/valgrind/logs/2009-10-21-19.42.log
> and broken out by test in
>  http://kegel.com/wine/valgrind/logs/2009-10-21-19.42/
> Please send me the suppression record for any warnings that are
> spurious or benign, and mention the reason for each.

Listview is among leaders, that's nice. Dan, do I need something special 
to build current
valgrind? Is there any special gcc/binutils requirements?
> valgrind/memcheck.h \
>     valgrind/valgrind.h

Will this be installed by valgrind install script (if it exists of course)?

[Dan Kegel]

On Thu, Oct 22, 2009 at 7:28 AM, Nikolay Sivov <bunglehead(a)gmail.com> wrote:
> Dan, do I need something special to
> build current valgrind? Is there any special gcc/binutils requirements?

I don't think there are any special tool requirements (other than
possibly "don't use binutils-gold", but that's true for wine as well).
You *do* need a small patch.  I've updated
  http://wiki.winehq.org/Wine_and_Valgrind
to give the procedure.
> >    valgrind/memcheck.h \
> >    valgrind/valgrind.h
>
> Will this be installed by valgrind install script (if it exists of course)?

"sudo make install" will install them.  (Though if you use --prefix, it will
install them there instead of /usr/include.  The wiki has complete
instructions.)
- Dan