December 2007 - Wine-devel - list.winehq.org

Re: msi: fix memory leaks (found by Smatch).
by James Hawkins 09 Dec '07

09 Dec '07

On Dec 9, 2007 4:44 AM, Lionel_Debroux <lionel_debroux(a)yahoo.fr> wrote: > "msi_build_insertsql_data" and "MSI_DatabaseImport" in > dlls/msi/database.c get flagged by Michael Stefaniuc's unfree-wine.pl > script. > > It looks there are more leaks than unfree-wine.pl reports, due to > constructs of the form > ptr = msi_realloc(ptr, size); > Fixed those as well. > > 2007-12-09 Lionel Debroux <lionel_debroux(a)yahoo.fr> > * dlls/msi/database.c: > msi: fix memory leaks (found by Smatch). > - columns = msi_realloc(columns, sql_size * sizeof(WCHAR)); - if (!columns) + temp_columns = msi_realloc(columns, sql_size * sizeof(WCHAR)); + if (!temp_columns) + { + HeapFree( GetProcessHeap(), 0, columns); return NULL; + } lstrcatW(columns, expanded); This is wrong. There's no guarantee that realloc returns the same position in memory. -- James Hawkins

1 0

Re: [PATCH 3/6] COMDLG32: printdlg: check hook pointer (resend 3)
by Dmitry Timoshkov 09 Dec '07

09 Dec '07

"Vitaly Lipatov" <lav(a)etersoft.ru> wrote: > --- dlls/commdlg/printdlg.c 11 Jan 2006 11:37:03 -0000 1.100 > +++ dlls/comdlg32/printdlg.c 24 Feb 2006 17:47:53 -0000 The way you've generated this patch looks pretty strange. > /* Call user paint hook if enable */ > if (pda->dlga->Flags & PSD_ENABLEPAGEPAINTHOOK) > - if (pda->dlga->lpfnPagePaintHook(hwndDlg, uMsg, wParam, lParam)) > + if (pda->dlga->lpfnPagePaintHook && pda->dlga->lpfnPagePaintHook(hwndDlg, uMsg, wParam, lParam)) You have to add the checks existing in PageSetupDlgA to PageSetupDlgW instead. Also, in order to increase the probability of your patches to be included should either point to existing bug that each of your patches fix, or/and add a test case which doesn't pass without your patch and does with it. -- Dmitry.

2 2

speeding up regression testing by disabling building of tests?
by Dan Kegel 09 Dec '07

09 Dec '07

Lots of the time in regression searches is spent compiling tests, but when you're only going to run the app that's breaking, that's not too useful. How does one disable compilation of conformance tests? I have this funny feeling Lei or Mikolaj did this recently, but I couldn't track it down.

2 2

Read Application Memory
by Kai Lauterbach 09 Dec '07

09 Dec '07

Hi, at the moment I'm writing a little Linux based program to read some variables from a Windows application. But my problem is, that i can't access to the virtual memory of the Win App running under wine. I don't know how to do this. c/c++ isn't the problem but the right way to do it. I've already read the wineserver's main.c and fd.c but I think that it's not a good way to put a code into, which give's me the chance to read something. I need sth to connect, like /dev/mem or a network port. My second idea was the wine debugger. Is it possible to get access the application memory at runtime from this? I mean, without stopping the application? I hope you can help me and I've described my problem as clear as possible ;) Greetings, Kai -- Kai Lauterbach <klaute(a)web.de>

1 0

Re: Alexander Nicolaysen Sørnes : comdlg32: PageSetupDlgA: Read margins from correct edit boxes.
by Vitaly Lipatov 09 Dec '07

09 Dec '07

Hello, I already posted the same patch to wine-patches one month ago: http://www.winehq.org/pipermail/wine-patches/2007-November/046156.html Why my patch was ignored? I am in some black list? В сообщении от 8 декабря 2007 Alexandre Julliard написал(a): > Module: wine > Branch: master > Commit: 51086b5513a436f51460d006f4fa7156e9d58a92 > URL: > http://source.winehq.org/git/wine.git/?a=commit;h=51086b5513a436f51460d006f >4fa7156e9d58a92 > > Author: Alexander Nicolaysen Sørnes <alex(a)thehandofagony.com> > Date: Sat Dec 8 13:53:32 2007 +0100 > > comdlg32: PageSetupDlgA: Read margins from correct edit boxes. > > --- > > dlls/comdlg32/printdlg.c | 4 ++-- > 1 files changed, 2 insertions(+), 2 deletions(-) > > diff --git a/dlls/comdlg32/printdlg.c b/dlls/comdlg32/printdlg.c > index e04e145..5baa90e 100644 > --- a/dlls/comdlg32/printdlg.c > +++ b/dlls/comdlg32/printdlg.c > @@ -2890,10 +2890,10 @@ PRINTDLG_PS_WMCommandA( > GETVAL(id, pda->curdlg.rtMargin.left); > break; > case edt5: > - GETVAL(id, pda->curdlg.rtMargin.right); > + GETVAL(id, pda->curdlg.rtMargin.top); > break; > case edt6: > - GETVAL(id, pda->curdlg.rtMargin.top); > + GETVAL(id, pda->curdlg.rtMargin.right); > break; > case edt7: > GETVAL(id, pda->curdlg.rtMargin.bottom); -- Vitaly Lipatov, ALT Linux Team Russia, Saint-Petersburg, www.etersoft.ru

2 2

Saturday's(!) valgrind results
by Dan Kegel 09 Dec '07

09 Dec '07

For the first time I can recall, Alexandre committed lots of changes on a Saturday, so I did a valgrind run. http://kegel.com/wine/valgrind/logs-2007-12-08-summary.txt shows the highlights; basically, lots more warnings like + Syscall param writev(vector[1]) points to uninitialised byte(s) For example, the one at the top of http://kegel.com/wine/valgrind/logs-2007-12-08/vg-user32_edit-diff.txt + Syscall param writev(vector[1]) points to uninitialised byte(s) + at (within /lib/ld-2.5.so) + by writev (in /lib/tls/i686/cmov/libc-2.5.so) + by wine_server_call (server.c:244) The only change in the server was http://source.winehq.org/git/wine.git/?a=commitdiff;h=edd3f264d95c0c6bed9c1… so I guess the new errors reflect additional error completion events being sent, and that kind of event just happens to contain some uninitialized bytes. So this is probably in Alexandre's court, and my guess is he'll consider this a false positive. - Dan -- Wine for Windows ISVs: http://kegel.com/wine/isv

1 0

Re: Valgrind support for our unit tests
by Robert Shearman 08 Dec '07

08 Dec '07

Dan Kegel wrote: > +{ > + todo_suppress_exception_addr4 > + Memcheck:Addr4 > + fun:__wine_call_from_32_regs > + obj:/home/dank/wine-git/dlls/ntdll/ntdll.dll.so > + obj:/home/dank/wine-git/dlls/ntdll/ntdll.dll.so > + fun:raise_exception I'm not sure this will work for other people unless they log on as dank :-) However, I don't think maintaining the valgrind suppressions file in tree is viable since most of the warnings come from system libraries which means that it is very dependent on user's distribution. It might be best to set up an external git tree for this, which would allow for people to submit patches for any bogus warnings in the runs that you publish on your website. -- Rob Shearman

4 4

File a bug report?
by Gilles May 08 Dec '07

08 Dec '07

Hi everybody. After updating WINE to from 0.9.49 to 0.9.50 I noticed major slowdown in Company of Heroes, to the point that I get like 1 Frame every 2 seconds. Wine spams the console with: fixme:d3d_shader:IWineD3DVertexShaderImpl_CompileShader Vertexshader 0x4d20b78 recompiled more than 50 times Trying to find out the cause for the slowdown I did my first attempt at using git bisect and I could trace the slowdown to this patch: 959212304dfaf0e7459374593b20b74fe22a0b57 is first bad commit commit 959212304dfaf0e7459374593b20b74fe22a0b57 Author: Stefan Dösinger <stefan(a)codeweavers.com> Date: Tue Nov 20 21:14:10 2007 +0100 wined3d: Track vertex declaration changes on vertex shaders. If an attribute has type D3DDECLTYPE_D3DCOLOR, the red and blue channels are swizzled in the shader. Since the attribute is stored in the vertex declaration and not the vertex shader, it can change by setting a new vertex declaration. If this happens, we have to recompile the shader with the swizzling of that specific attribute turned on or off. :040000 040000 327661a4d0f8a02bdc18380084589c891b64af9c e35a2bcbf9514e4f261d8b7a93dcb24c0b3c89a7 M dlls Now without that patch there is slight texture corruption in the game, with the patch it's much much less but it renders the game unplayable. The reason I didn't file a bug report is that I am not sure if it is really a bug or not since obviously the patch fixes in-game texture corruption. Regards, Gilles

3 2

Re: ddraw: fix wrong initialization order in IDirect3DDevice vtable
by Stefan Dösinger 08 Dec '07

08 Dec '07

Am Sonntag, 9. Dezember 2007 03:00:28 schrieb Alexander Dorofeyev: > Helps bug#7584 (http://bugs.winehq.org/show_bug.cgi?id=7584) > --- > dlls/ddraw/device.c | 2 +- > 1 files changed, 1 insertions(+), 1 deletions(-) Looks correct to me, good catch!

1 0

Re: dlls/comctl32/listview.c warning elimination
by Alexandre Julliard 08 Dec '07

08 Dec '07

Gerald Pfeifer <gerald(a)pfeifer.com> writes: > Not a patch I am particularily proud of, but the best way I found to get > rid of > > listview.c:5043: warning: 'strW' might be used uninitialized in this function > > issued by GCC, and apart from the added cast it actually is simpler than > the original. It was done the way it is precisely to avoid the non-const cast, which triggers a warning too. -- Alexandre Julliard julliard(a)winehq.org

2 2