New valgrind results for Oct 26, 2009

View: New views
2 Messages — Rating Filter:   Alert me  

New valgrind results for Oct 26, 2009

by Dan Kegel-2 :: Rate this Message:

Reply to Author | View Threaded | Show Only this Message

http://kegel.com/wine/valgrind/logs/2009-10-26-08.26/
has current results.  I've tweaked the diff-* files
to be brutally minimalist, so if there's a diff- file,
it's more likely to be interesting.  On the downside,
you have to look at the non-diff file to see line numbers.

One of these days I've got to make these clickable...

Highlights:

"New" errors:

http://kegel.com/wine/valgrind/logs/2009-10-26-08.26/diff-d3d10_effect.txt
has a new warning,
see http://www.winehq.org/pipermail/wine-devel/2009-October/079467.html

http://kegel.com/wine/valgrind/logs/2009-10-26-08.26/diff-mshtml_htmllocation.txt
shows a bunch of "new" warnings, see
http://bugs.winehq.org/show_bug.cgi?id=20485

Fixed:

http://kegel.com/wine/valgrind/logs/2009-10-26-08.26/diff-gdiplus_image.txt
shows that Kirill's patch did the trick
http://kegel.com/wine/valgrind/logs/2009-10-26-08.26/diff-oleaut32_tmarshal.txt
shows that Jeremy's patch helped
http://kegel.com/wine/valgrind/logs/2009-10-26-08.26/diff-urlmon_url.txt and
http://kegel.com/wine/valgrind/logs/2009-10-26-08.26/diff-wininet_http.txt
seem to show benefits from Hans' patch

Thanks to everyone who fixed memory problems so far.  There's lots
more to do, check
out your favorite module and see what dirt valgrind has on it!



Parent Message unknown Re: New valgrind results for Oct 26, 2009

by André Hentschel :: Rate this Message:

Reply to Author | View Threaded | Show Only this Message

Dan Kegel schrieb:

> 2009/10/28 André Hentschel <nerv@...>:
>> I think i found a false postitv. its http://kegel.com/wine/valgrind/logs/2009-10-26-08.26/vg-ntdll_env.txt
>> the code in ntdll/tests/env.c looks like:
>>        ...
>>        name.Buffer = bn;
>>        ...
>>        value.Buffer = bv;
>>        ...
>>        nts = pRtlQueryEnvironmentVariable_U(small_env, &name, &value);
>>        ...
>>        if(memcmp(bv, bn, test->len*sizeof(WCHAR)) == 0)
>>
>> So valgrind thinks bn and bv are uninitialised, but thats wrong.
>> Do you agree with me?
>
> I actually looked at this one for an hour the other day.
> I think valgrind's probably right, but the test is confusingly written.

So i send this issue to wine-devel to know what other devs think about it.

--

Best Regards, André Hentschel