March 2005 - Wine-devel - list.winehq.org

Re: Add new "nrelay" debug channel - useful for debugging GPFs in target applications
by Andreas Mohr 02 Mar '05

02 Mar '05

Hi, On Wed, Mar 02, 2005 at 04:09:36PM +1100, Troy Rollo wrote: > The nrelay ("native relay") debug channel acts to limit the output of the > relay debug channel, and only works if the relay debug channel is also on. > When "nrelay" is on, "relay" will not log calls made from one built-in > library to another built-in library, or within a built-in library. Thus each > call logged is a call made directly by the target application or by one of > the native DLLs it depends on. Yay!! Thanks a lot! This one is VERY important, since developer confusion by overwhelmingly big and mostly less useful relay logs in recent years has grown a lot, I'd say. One suggestion: maybe make *only* logging external relays the default, and add a wrelay channel to add back the logging of Wine inter-dll calls? This might make more sense even, since you're often not interested at all in Wine interactions, but what the target app does. Any suggestions about the average percentage where people are interested in app-only calls as opposed to a full relay log? That would then tell us which way to implement this limiting. I'm not sure about the percentage, since I haven't done much development in recent years. Andreas Mohr

2 1

Re: USER32: Sysparams unit tests fixes and additions
by Paul Vriens 02 Mar '05

02 Mar '05

On Tue, 2005-03-01 at 05:53, jchevrier(a)nexicom.net wrote: > Changelog: > > * Created SPI_{GET,SET}KEYBOARDPREF and SPI_{GET,SET}SCREENREADER unit tests > * Removed a trace that shouldn't have been there > * Add return checking to SPI_{GET,SET}WHEELSCROLLLINES and > SPI_{GET,SET}MENUSHOWDELAY unit tests so they no longer run on Windows 95 > where they aren't supported > * Changed minimum hovertime in SPI_{GET,SET}MOUSEHOVERTIME to 10 as Windows XP > defaults to 10 any value below that Hi, the extra trace was my doing :-(. I've added this and after that I realized that SPI_{GET|SET} doesn't clean the LastError after a correct call. That's why I think it's necessary to do a SetLastError(0) before every check of GetLastError. That's the reason I had these failures on win98 and SPI_GETICONTITLEWRAP. The Error was still set by one of the previous called functions (I think a ..W call, the ..A call was fine). That still leaves us with the questions, whether it would be good to check for ERROR_CALL_NOT_IMPLEMENTED after each call? Cheers, Paul Vriens.

2 1

Re: [wined3d] texture interface separation.
by Oliver Stieber 02 Mar '05

02 Mar '05

> wrote: > > > > Hi, > > > > This patch start to separate the interfaces > in > > > > wine3d3. Complete separation requires > > > queryInterface > > > > support which I'll send through soon, this > patch > > > has > > > > no dependencies but provides the final > interface > > > > needed by later revisions. > > > > > > > Hi Oliver, > > > > > > this patch couldn't be applied cleanly to CVS. > Just > > > to let you know, I > > > didn't check further. > > > > > > I'm wondering if AJ will have many comments on > the > > > patches. I hope not > > > as your work is going to make a lot of people > happy. > > It's a diff against CVS. But I think you need a diff > against CVS + your > previous patch. Or am I missing something ? > The patches aren't incremental, that way one can fail but it shouldn't affect the rest being merged. I have another couple of discrete patches to send in then the patches become dependant on some of the interface changes, but hopefully not the underlying functionality changes. I also don't have enough HDD space to keep incremental versions and check that they compile/work especially with all the interface/include changes. If anyone knows of a good tool for managing patches let me know (not cerversa!), there's a surprising lack of tools available. at the moment I'm using the following... #!/bin/bash #given a file, replace the first path with a new path #diff the two files stripedpath=$(echo $1 | grep -o "/.*") kdiff3 $1 "$2/$stripedpath"& kwrite $1 & kwrite "$2/$stripedpath"& -- diffdueo clean/wine/.... working Very 21st century. A tool for merging would be good too, e.g. merge patch1.patch patch2.patch > patch_merged.patch. if one doesn't exist I'll write it and put it on my site. Send instant messages to your online friends http://uk.messenger.yahoo.com

2 6

Re: [PATCH] Added 2 functions to shell32
by Jacek Caban 01 Mar '05

01 Mar '05

> I doubt shell32 function on windows does this check, also original windows shell32 > function is a lot more complicated and longer, I've compared the results between > real and my code and they are correct. I wrote a small test for this function showing that your implementation is almost correct. It failes in two cases: 1. If a path ends with '\' and is compared with one that doesn't, then it's length is less than the other one, but both are the same paths. 2. If a path doesn't end with '\' and is compared with a longer one that has the same first len chars, but is not the same, ex: c:\test c:\test123 Changelog: Added test for PathIsEqualOrSubFolder Index: dlls/shell32/tests/shellpath.c =================================================================== RCS file: /home/wine/wine/dlls/shell32/tests/shellpath.c,v retrieving revision 1.6 diff -u -p -r1.6 shellpath.c --- dlls/shell32/tests/shellpath.c 10 Feb 2005 19:19:35 -0000 1.6 +++ dlls/shell32/tests/shellpath.c 28 Feb 2005 21:19:31 -0000 @@ -871,6 +871,55 @@ static void testNonExistentPath(void) userShellFolders); } +typedef BOOL (WINAPI*fnPathIsEqualOrSubFolder)(LPCWSTR path1, LPCWSTR path2); + +static void test_PathIsEqualOrSubFolder() +{ + fnPathIsEqualOrSubFolder pPathIsEqualOrSubFolder; + BOOL ret; + static const WCHAR path1[] = {'c',':','\\','t','e','s','t',0}; + static const WCHAR path2[] = {'c',':','\\','t','e','s','t','\\','t','e','s','t',0}; + static const WCHAR path3[] = {'c',':','\\','t','e','s','t','1','2','3',0}; + static const WCHAR path4[] = {'c',':','\\','t','e','s','t','\\',0}; + static const WCHAR path5[] = {'c',':','\\','t','e','s','t','\\','t','e','s','t','2', + '\\','.','.','\\','t','e','s','t',0}; + + pPathIsEqualOrSubFolder = GetProcAddress(hShell32, (LPSTR)755); + if(!pPathIsEqualOrSubFolder) + return; + +#define TEST_PATH(p1, p2, r) \ + ret = pPathIsEqualOrSubFolder(p1, p2); \ + ok(ret == r, "PathIsEqualOrSubFolder(" #p1 " " #p2 ") failed, expected: %x\n", r); + + TEST_PATH(path1, path1, TRUE); + TEST_PATH(path1, path2, TRUE); + TEST_PATH(path1, path3, FALSE); + TEST_PATH(path1, path4, TRUE); + TEST_PATH(path2, path1, FALSE); + TEST_PATH(path2, path2, TRUE); + TEST_PATH(path2, path3, FALSE); + TEST_PATH(path2, path4, FALSE); + TEST_PATH(path3, path1, FALSE); + TEST_PATH(path3, path2, FALSE); + TEST_PATH(path3, path3, TRUE); + TEST_PATH(path3, path4, FALSE); + TEST_PATH(path4, path1, TRUE); + TEST_PATH(path4, path2, TRUE); + TEST_PATH(path4, path3, FALSE); + TEST_PATH(path4, path4, TRUE); + TEST_PATH(path5, path1, FALSE); + TEST_PATH(path5, path2, FALSE); + TEST_PATH(path5, path3, FALSE); + TEST_PATH(path5, path4, FALSE); + TEST_PATH(path5, path5, TRUE); + TEST_PATH(NULL, path1, FALSE); + TEST_PATH(path1, NULL, FALSE); + TEST_PATH(NULL, NULL, FALSE); + +#undef TEST_PATH +} + START_TEST(shellpath) { if (!init()) return; @@ -892,5 +941,6 @@ START_TEST(shellpath) testWinDir(); testSystemDir(); testNonExistentPath(); + test_PathIsEqualOrSubFolder(); } }

3 2

New uninstaller
by Jonathan Ernst 01 Mar '05

01 Mar '05

As requested, I'm trying to make a new uninstaller using a dialog instead of a window. It's my first windows program and I have something working but there are still some issues. If someone could help with some of the following problems, I'd be very thanfull. 1) Strange problem with UpdateWindow uninstaller.c, line 264 I implemented a search/filter field and I want to repaint the window after each input in the EDITTEXT control. However, when I call UpdateWindow on this line, the value returned is 1 so that a WM_PAINT message should be sent right ? But strangly no message is sent and I have to make something else (like moving the dialog) to force the window to repaint and filter the entries. The funny thing is that UpdateWindow on line 279 works just fine. I guess it's a bug in my code and not in Wine so I'd be glad if you could help me. 2) Improve filter code Currently I filter the entries that match exactly the filter string. I would like to at least make a case insensitive strstr on the string. What's the best way to do it (See uninstaller.c, line 235) 3) Stringtables a)Is there a way to load localized strings without having to guess beforehand the size of the string in each language ? b)Is there a way to not call LoadString for each string ? c)Where is the best place to load every string of a stringtable ? in main(), just before calling DialogBox() ? 4) Please tell me if you see things that are bad in the attached code, so that I'll fix it before submitting a patch. Thanks you for your help

7 13

Regression in Rebar
by Maxime Bellengé 01 Mar '05

01 Mar '05

Hello This patch is causing a regression in emule, the rebar simply does not display. a+ Maxime Modified files: dlls/comctl32 : rebar.c Log message: Robert Shearman <rob(a)codeweavers.com> - Make row number be zero-based. - Improvements to dumping functions to not dump out fields that may not have been filled in.

2 4

Tools for identifying what a DLL provides?
by David D. Hagood 01 Mar '05

01 Mar '05

I have a DLL from a manufacturer that I don't have the source for, nor do I have the header or the import library for. Since Wine developers frequently have to reverse engineer what a given DLL provides, does anybody have a good recommendation for a tool that can scan the DLL and report what functions it exports, and their arg types and return types?

4 4

Re: [wined3d] texture interface separation.
by Oliver Stieber 01 Mar '05

01 Mar '05

2 1

Re: Wine cabinet.dll FDI Conformace Test Patch
by Daniel Kegel 01 Mar '05

01 Mar '05

Andreas wrote: > It's an unspoken rule that preprocessor directives (#ifdef, ...) > always start at the very beginning of a line to make sure one recognizes > immediately which parts of the code are being processed and which ones aren't. Good point. I should have caught that style violation when I reviewed their code for them this weekend. > AFAICS you're not using wine test suite infrastructure, e.g. ok() etc. > > Using printf() directly strikes me as being quite bad, too... I think they're using the wine test infrastructure properly everywhere. The only exception is the disabled self-test in tvfs.c, but I think that's fine. It's not even built, it was just used while developing the test harness. > Apart from that it appears to be very good. Cool! - Dan

2 1

Re: winedbg questions
by Pouech Eric DMI AEI CAEN 01 Mar '05

01 Mar '05

> I'm trying to use winedbg to run the direct sound regression > tests and have two questions: > > Execution stops when starting the first dsound.dll thread. Is > this normal? Is the stack trace with no symbol information > normal? no, because of this in your trace Couldn't initiate DbgHelp the only reason this fails is because dbghelp cannot get to the wine-loader in the running programm try to set the env var WINELOADER to the {pk}loader (full path) you're using and check again A+

1 0