April 2008 - Wine-Devel - list.winehq.org

Abusing valgrind to find reference leaks
by Maarten Lankhorst April 29, 2008

April 29, 2008

Hi all, I've managed to abuse valgrind to show backtraces of all places where an AddRef and Release is called when there is a memory, this is superior to wading through an endless pile of logs without even guarantee it will work. I thought I should share it for the following reasons: 1. Valgrind is an awesome tool 2. Reference leaks are usually impossible or very hard to find by using debug logs All I'm doing is leaking memory until the destructor is called, which gives quite useful results, I had to add a exit(1) after the leak occured because the samples get destroyed regardless when the interface controlling those samples is released. The log was created with: valgrind --trace-children=yes --leak-check=full --show-reachable=yes --leak-resolution=high -- wine quartz_test.exe.so filtergraph 2>&1 | tee valgrind.log And then I removed all the memory leaks and warnings not related to AddRef and Release. Ideally I could get it to automatically sort the outputted data based on the order of calling, but I'm not 100% sure how to do that yet. I could do 1,000,000 + GetTickCount(), but the results are usable now regardless. This was useful for me to get the backtraces of all _AddRef's and _Release's *ONLY* for the leaking interfaces, which saves a lot of time determining whether the specific interface is leaking and not, and when it starts leaking, and then determining who calls and where it is called, so I hope it is useful for you too! Cheers, Maarten.

1 0

Re: widl: callback, code, comm_status and in_line are attribute names, not keywords.
by Robert Shearman April 29, 2008

April 29, 2008

Robert Shearman wrote: > --- > tools/widl/parser.l | 8 ++++---- > 1 files changed, 4 insertions(+), 4 deletions(-) Just to note, this patch depends on the patch series that I sent yesterday and that I'm about to resend now. -- Rob Shearman

1 0

re: 1.0 idea - ntdll - add a messagebox for missing dlls
by Dan Kegel April 29, 2008

April 29, 2008

Hmm, maybe make explorer responsible for putting up the messagebox, and set up some IPC method (good lord, not COM, just some mailbox or something) for the loader to notify it...

6 26

Bad recommended package lists
by Francois Gouget April 29, 2008

April 29, 2008

http://wiki.winehq.org/Recommended_Packages The Wiki's Recommended Packages list contains a lot of incorrect recommendations. I have already cleaned up the Debian section, but there seems to be a lot of copy/paste that spread the bad information to the RPM distros. So we'd need volunteers for pruning and updating these lists for Fedora and RHEL / WBEL. Here's a list of the suspicious or downright bad package recommendations I spotted: * audiofile-devel This is a library for accessing audio files. Wine does not use it. Maybe it was confused with AudioIO (which is not used on Linux anyway). * e2fsprogs-devel Why would Wine need ext2fs tools? * expat-devel We use libxml2, not expat. * glib2-devel That's not the glibc. That's a Gnome library which Wine never used. * isdn4k-utils-devel Maybe this is for libcapi20? * krb5-devel As far as I know we don't use Kerberos in Wine (or it's experimental code that is not enabled in the mainline). * libao-devel Wine does not use libao. Instead it has backends for most of the audio systems out there, although there's always a new kid on the block that's missing. * libmng-devel That's for MNG files, the moving version of PNG files. But what we use is libpng, not libmng. * libusb-devel Wine did use libusb at one point but now it's using libhal. * libogg-devel * libvorbis-devel We don't use those. * libxml, libxml-devel We use libxml2, not the deprecated libxml1. Now, maybe some distributions use 'libxml-devel' as the package name for libxml2 but this should be clarified, especially since libxml2-devel is mentionned too. * qt-devel That's QT, the toolkit used by KDE. Wine has never used it. * libXxf86dga-devel As far as I can tell we don't use libXxf86dga anymore. * opencdk-devel, libgpg-error-devel, libgcrypt-devel These are all related to GPG. We have nothing to do with GPG so we don't need them. * libXft-devel Wine uses FreeType and fontconfig but not libXft as far as I know. * gnutls-devel I believe we used GNU TLS at some point but now we use OpenSSL. * cyrus-sasl-devel Cyrus SASL also has to do with cryptography but as far as I can tell we never used it. * libieee1284-devel For parallel port access. Not used as far as I can tell. * libXmu-devel Contains 'miscellaneous utility convenience functions for X libraries to use'. I don't think Wine uses it. -- Francois Gouget <fgouget(a)free.fr> http://fgouget.free.fr/ RFC 2549: ftp://ftp.isi.edu/in-notes/rfc2549.txt IP over Avian Carriers with Quality of Service

3 3

Re: [resend] [Try 4] [1/2] cabinet.dll -> FDICopy: Test added
by James Hawkins April 29, 2008

April 29, 2008

2008/4/29 Vitaly Perov <vitperov(a)etersoft.ru>: > jhawkins(a)codeweavers.com : cabinet: Revert "cabinet: Fix for FDICopy with an > empty cabinet file." > > Hi, > > > > The test was incorrect and failed on all platforms. Vitaly, please > > ask someone to test your patches on a real windows system in the > > future if you don't have access to one. > > I have tested this patch on Win98. It didn't failed. > But on WinXP it failed. > > The pach comes with this test fix a problem, so installation of certain > software (it includes one empty cab-file ) doesn't fail anymore. > I don't think, that this patch can cause any problem with a software developed > for WinXP. > > So, please look it trough once more. > Just look at the test results: http://test.winehq.org/data/200804252248/ It fails on every single platform, including win98. -- James Hawkins

1 0

Re: user32: Send WM_IME_SETCONTEXT message when the window is focusedor unfocused.
by Dmitry Timoshkov April 29, 2008

April 29, 2008

"ByeongSik Jeon" <bsjeon(a)hanmail.net> wrote: > --- a/dlls/user32/tests/msg.c > +++ b/dlls/user32/tests/msg.c > @@ -10156,7 +10156,7 @@ static const struct message wm_lb_click_0[] = > { HCBT_SETFOCUS, hook }, > { WM_KILLFOCUS, sent|parent }, > { WM_IME_SETCONTEXT, sent|wparam|optional|parent, 0 }, > - { WM_IME_SETCONTEXT, sent|wparam|optional, 1 }, > + { WM_IME_SETCONTEXT, sent|wparam|defwinproc|optional, 1 }, Did you test this change under Windows? At the time I've added this test it passed cleanly under up to date XP SP2. -- Dmitry.

2 3

Re: Alternative for isinf (fix for bug 12008) (version 4)
by Stefan Dösinger April 29, 2008

April 29, 2008

1 0

Re: [PATCH 1/4] widl: Keep const attributes applied to pointers when writing out the type.
by Alexandre Julliard April 29, 2008

April 29, 2008

Robert Shearman <rob(a)codeweavers.com> writes: > Use an attribute to store the const qualifier for the pointer and type. > > Allow multiple type-qualifiers to be applied to a type by adding a > declaration-specifier rule that encompasses type-qualifiers and types. > --- > tools/widl/header.c | 5 ++- > tools/widl/parser.y | 103 > ++++++++++++++++++++++++++++++++++++++++------- > tools/widl/widltypes.h | 2 +- > 3 files changed, 92 insertions(+), 18 deletions(-) > > This is a resend of a previous patch, with the bug that Alexandre > spotted fixed. The bug was that typedef'd types and arrays would no > longer have any const modifier applied to them. That works now, though there is still something not quite right: --- include/d3d10.h.old 2008-04-29 13:38:38.000000000 +0200 +++ include/d3d10.h 2008-04-29 13:39:08.000000000 +0200 @@ -3455,12 +3455,12 @@ virtual void STDMETHODCALLTYPE VSSetConstantBuffers( UINT StartSlot, UINT NumBuffers, - ID3D10Buffer **ppConstantBuffers) = 0; + ID3D10Buffer **const ppConstantBuffers) = 0; It should be "ID3D10Buffer * const *" instead. -- Alexandre Julliard julliard(a)winehq.org

2 1

Re: user32: Post the WM_KEY{DOWN|UP} message when WM_IME_KEY{DOWN|UP}message is generated by IME.
by Dmitry Timoshkov April 29, 2008

April 29, 2008

"ByeongSik Jeon" <bsjeon(a)hanmail.net> wrote: > The PeekMessage, GetMessage CAN CATCH this WM_KEYDOWN message. > I think this problem don't need the test code. > This problem is very simple case and clearly. This really needs the tests. First to show that this change is really correct, second to prevent regressions in future when someone else will try to change it. -- Dmitry.

1 0

Re: 3/2 WineD3D: add D3DPRASTERCAPS_MULTISAMPLE_TOGGLE support
by Stefan Dösinger April 29, 2008

April 29, 2008

Am Montag, 28. April 2008 22:08:28 schrieb Roderick Colenbrander: > Hi, > > Add a missing d3d9 capability. For now add it behind a dxVersion check. We > should get rid of those after Wine 1.0 but right now there is no filtering > code in d3d8, so don't risk any regressions now. IMHO adding a caps filter to dx8 is less risky than adding multisampling support, so I don't see a reason not to do that and avoid the dxVersion check

4 3