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
2009/10/22 Dan Kegel dank@kegel.com:
- 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.
Ben Klein wrote:
2009/10/22 Dan Kegel dank@kegel.com:
- 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
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)?
On Thu, Oct 22, 2009 at 7:28 AM, Nikolay Sivov bunglehead@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
-
Ben Klein -
Dan Kegel -
Michael Stefaniuc -
Nikolay Sivov