October 2001 - Wine-Devel - list.winehq.org

Nasty Evil Memory Fragmentation fix
by Gavriel State Oct. 10, 2001

Oct. 10, 2001

We just spent about 3 days tracking down a very subtle memory fragmentation problem in an app we're working on. The problem: App allocates lots of little blocks, then a few large blocks, then lots of little blocks. Throughout this process, many blocks would be freed from all over the heap. Repeat ad infinitum. We ended up creating a new subheap every time one of these larger blocks came along. The new space from the subheap then went to the top of the freelist, so all the small allocations then filled it up until the next large block allocation. IE: We never got a chance to reuse freed blocks in older subheaps - at least not very well. This fix addresses the problem by pushing 'large' freed blocks to the end of the free list, leaving 'small' blocks at the front, to be found sooner. The size of 'large' vs 'small' is tunable. Despite the workaround, we're still not too pleased with the current heap allocator. It's quite slow, and still not as efficient as it could be. It would probably be worth the effort to integrate a new allocator - anyone know if there's a high-quality Wine-license- compatible allocator out there? -Gav -- Gavriel State, CEO TransGaming Technologies Inc. http://www.transgaming.com gav(a)transgaming.com --- heap.c Wed Sep 26 19:11:21 2001 +++ /home/gavriels/transgaming/wine/memory/heap.c Tue Oct 9 23:53:15 2001 @@ -59,6 +59,8 @@ #define QUIET 1 /* Suppress messages */ #define NOISY 0 /* Report all errors */ +#define FREE_BLOCK_INSERTION_THRESHOLD 4096 + #define HEAP_NB_FREE_LISTS 4 /* Number of free lists */ /* Max size of the blocks on the free lists */ @@ -229,11 +231,11 @@ /*********************************************************************** - * HEAP_InsertFreeBlock + * HEAP_InsertFreeBlockAtFront * - * Insert a free block into the free list. + * Insert a free block into the front of the free list. */ -static void HEAP_InsertFreeBlock( HEAP *heap, ARENA_FREE *pArena ) +static void HEAP_InsertFreeBlockAtFront( HEAP *heap, ARENA_FREE *pArena ) { FREE_LIST_ENTRY *pEntry = heap->freeList; while (pEntry->size < pArena->size) pEntry++; @@ -244,6 +246,31 @@ pEntry->arena.next = pArena; } +/*********************************************************************** + * HEAP_InsertFreeBlockAtEnd + * + * Insert a free block into the end of the free list. + */ +static void HEAP_InsertFreeBlockAtEnd( HEAP *heap, ARENA_FREE *pArena ) +{ + FREE_LIST_ENTRY *pEntry = heap->freeList; + + /* Determine which free list contains blocks of the appropriate size. */ + while (pEntry->size < pArena->size) pEntry++; + + /* But we want to put this block on the end of said appropriate free list. + * So instead, find the list for the next size up, knowing that they are + * all connected. */ + if (++pEntry == heap->freeList+HEAP_NB_FREE_LISTS) + pEntry = heap->freeList; + + pArena->size |= ARENA_FLAG_FREE; + pArena->next = &pEntry->arena; + pArena->prev = pEntry->arena.prev; + + pArena->next->prev = pArena; + pArena->prev->next = pArena; +} /*********************************************************************** * HEAP_FindSubHeap @@ -372,7 +399,11 @@ /* Last, insert the new block into the free list */ pFree->size = size - sizeof(*pFree); - HEAP_InsertFreeBlock( subheap->heap, pFree ); + + if ((pFree->size & ARENA_SIZE_MASK) > FREE_BLOCK_INSERTION_THRESHOLD) + HEAP_InsertFreeBlockAtEnd( subheap->heap, pFree ); + else + HEAP_InsertFreeBlockAtFront( subheap->heap, pFree ); }

2 1

Fwd: Re: Remove annoying fatal_error() after wineserver crash
by Malte Starostik Oct. 10, 2001

Oct. 10, 2001

Seems I'm too used to mailing lists which set Reply-to to the list... Am Wednesday 10 October 2001 01:15 schrieben Sie: > Malte Starostik <malte(a)kde.org> writes: > > not sure about the reason for that "the socket exists, but I can't > > connect" message, but IMHO it's pretty annoying. I hit this fatal_exit() > > very often while debugging reaktivate. And in such cases - or when > > starting an app from a desktop / startmenu icon - when wine's STDERR goes > > to ~/.xsession-errors instead of a shell, the user has no chance to > > figure why an app won't start. Well, long story, short patch which Works > > for Me (TM) but I don't know if there's anything wrong with doing this... > > The problem with this is that it creates a race condition and we may > end up with two servers running in the same directory, which is bad. > > The error message should only happen when a previous server has > crashed, which is a critical error (the server should never crash no > matter what the app does). In that case you should also have a core > file in the same directory as the socket, which you can use to debug > and fix the crash. Thanks for the hint, I'll try to reproduce it with an appropriate ulimit set. But right, an actual wineserver crash didn't happen in a while. Unfortunately I have some system instability currently, so many times the message was caused by a stale socket lying around after a solid lockup of my box. What tends to happen rather often though is that wineserver keeps running after I quit all windows/winelib apps and consumes 100% CPU. In some of such cases, it'll only die on SIGKILL, which also leaves a socket around. I agree that neither a system lockup nor a hanging wineserver should happen though :) Maybe there could be a wineserver.pid file in the same dir and only if the process identified therein doesn't exist anymore, the socket will be autoremoved? But after all it's not too big a problem, guess there are more important issues. -- Malte Starostik PGP: 1024D/D2F3C787 [C138 2121 FAF3 410A 1C2A 27CD 5431 7745 D2F3 C787]

2 1

Re: Remove annoying fatal_error() after wineserver crash
by Alexandre Julliard Oct. 9, 2001

Oct. 9, 2001

Malte Starostik <malte(a)kde.org> writes: > not sure about the reason for that "the socket exists, but I can't connect" > message, but IMHO it's pretty annoying. I hit this fatal_exit() very often > while debugging reaktivate. And in such cases - or when starting an app from > a desktop / startmenu icon - when wine's STDERR goes to ~/.xsession-errors > instead of a shell, the user has no chance to figure why an app won't start. > Well, long story, short patch which Works for Me (TM) but I don't know if > there's anything wrong with doing this... The problem with this is that it creates a race condition and we may end up with two servers running in the same directory, which is bad. The error message should only happen when a previous server has crashed, which is a critical error (the server should never crash no matter what the app does). In that case you should also have a core file in the same directory as the socket, which you can use to debug and fix the crash. -- Alexandre Julliard julliard(a)winehq.com

1 0

Re: SHGetSpecialFolderPathA
by Chris Green Oct. 9, 2001

Oct. 9, 2001

That's the last time _I_ send a patch in the early hours of the morning! I missed a few commas... This one is better... Index: include/shlobj.h =================================================================== RCS file: /home/wine/wine/include/shlobj.h,v retrieving revision 1.47 diff -u -3 -p -r1.47 shlobj.h --- include/shlobj.h 2000/09/26 00:00:59 1.47 +++ include/shlobj.h 2001/10/09 21:53:39 @@ -461,6 +461,9 @@ HRESULT WINAPI SHGetSpecialFolderLocatio #define CSIDL_SENDTO 0x0009 #define CSIDL_BITBUCKET 0x000a #define CSIDL_STARTMENU 0x000b +#define CSIDL_MYDOCUMENTS 0x000c +#define CSIDL_MYMUSIC 0x000d +#define CSIDL_MYVIDEO 0x000e #define CSIDL_DESKTOPDIRECTORY 0x0010 #define CSIDL_DRIVES 0x0011 #define CSIDL_NETWORK 0x0012 @@ -473,6 +476,7 @@ HRESULT WINAPI SHGetSpecialFolderLocatio #define CSIDL_COMMON_DESKTOPDIRECTORY 0x0019 #define CSIDL_APPDATA 0x001a #define CSIDL_PRINTHOOD 0x001b +#define CSIDL_LOCAL_APPDATA 0x001c #define CSIDL_ALTSTARTUP 0x001d #define CSIDL_COMMON_ALTSTARTUP 0x001e #define CSIDL_COMMON_FAVORITES 0x001f @@ -494,8 +498,17 @@ HRESULT WINAPI SHGetSpecialFolderLocatio #define CSIDL_COMMON_ADMINTOOLS 0x002f #define CSIDL_ADMINTOOLS 0x0030 #define CSIDL_CONNECTIONS 0x0031 +#define CSIDL_COMMON_MUSIC 0x0035 +#define CSIDL_COMMON_PICTURES 0x0036 +#define CSIDL_COMMON_VIDEO 0x0037 +#define CSIDL_RESOURCES 0x0038 +#define CSIDL_RESOURCES_LOCALIZED 0x0039 +#define CSIDL_COMMON_OEM_LINKS 0x003a +#define CSIDL_CDBURN_AREA 0x003b +#define CSIDL_COMPUTERSNEARME 0x003d #define CSIDL_FOLDER_MASK 0x00ff - +#define CSIDL_FLAG_PER_USER_INIT 0x0800 +#define CSIDL_FLAG_NO_ALIAS 0x1000 #define CSIDL_FLAG_DONT_VERIFY 0x4000 #define CSIDL_FLAG_CREATE 0x8000 Index: dlls/shell32/shellpath.c =================================================================== RCS file: /home/wine/wine/dlls/shell32/shellpath.c,v retrieving revision 1.56 diff -u -3 -p -r1.56 shellpath.c --- dlls/shell32/shellpath.c 2001/08/16 18:49:57 1.56 +++ dlls/shell32/shellpath.c 2001/10/09 21:53:41 @@ -660,7 +660,7 @@ static const CSIDL_DATA CSIDL_Data[] = "Desktop", "Desktop" }, - { /* CSIDL_INTERNET (??) */ + { /* CSIDL_INTERNET */ 0, 1, /* FIXME */ NULL, NULL, @@ -715,22 +715,22 @@ static const CSIDL_DATA CSIDL_Data[] = "Start Menu", "Start Menu" }, - { /* not known */ - 0, 0, + { /* CSIDL_MYDOCUMENTS */ + 0, 1, /* FIXME */ NULL, NULL, }, - { /* not known */ - 0, 0, + { /* CSIDL_MYMUSIC */ + 0, 1, /* FIXME */ NULL, NULL, }, - { /* not known */ - 0, 0, + { /* CSIDL_MYVIDEO */ + 0, 1, /* FIXME */ NULL, NULL, }, - { /* not known */ + { /* unassigned */ 0, 0, NULL, NULL, @@ -795,8 +795,8 @@ static const CSIDL_DATA CSIDL_Data[] = "PrintHood", "PrintHood" }, - { /* not known */ - 0, 0, + { /* CSIDL_LOCAL_APPDATA */ + 0, 0, /* FIXME */ NULL, NULL, }, @@ -904,6 +904,66 @@ static const CSIDL_DATA CSIDL_Data[] = 0, 1, /* FIXME */ NULL, NULL + }, + { /* unassigned 32*/ + 0, 0, + NULL, + NULL, + }, + { /* unassigned 33*/ + 0, 0, + NULL, + NULL, + }, + { /* unassigned 34*/ + 0, 0, + NULL, + NULL, + }, + { /* CSIDL_COMMON_MUSIC */ + 0, 0, /* FIXME */ + NULL, + NULL, + }, + { /* CSIDL_COMMON_PICTURES */ + 0, 0, /* FIXME */ + NULL, + NULL, + }, + { /* CSIDL_COMMON_VIDEO */ + 0, 0, /* FIXME */ + NULL, + NULL, + }, + { /* CSIDL_RESOURCES */ + 0, 0, /* FIXME */ + NULL, + NULL, + }, + { /* CSIDL_RESOURCES_LOCALIZED */ + 0, 0, /* FIXME */ + NULL, + NULL, + }, + { /* CSIDL_COMMON_OEM_LINKS */ + 0, 0, /* FIXME */ + NULL, + NULL, + }, + { /* CSIDL_CDBURN_AREA */ + 0, 0, /* FIXME */ + NULL, + NULL, + }, + { /* unassigned 3C */ + 0, 0, + NULL, + NULL, + }, + { /* CSIDL_COMPUTERSNEARME */ + 0, 0, /* FIXME */ + NULL, + NULL, } }; #undef HKCU @@ -926,7 +986,7 @@ BOOL WINAPI SHGetSpecialFolderPathA ( TRACE("0x%04x,%p,csidl=%lu,0x%04x\n", hwndOwner,szPath,csidl,bCreate); - if ((folder > CSIDL_CONNECTIONS) || (CSIDL_Data[folder].hRootKey == 0)) + if ((folder > CSIDL_COMPUTERSNEARME) || (CSIDL_Data[folder].hRootKey == 0)) { ERR("folder unknown or not allowed\n"); return FALSE; _________________________________________________________ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com

1 0

Re: Service Pack Information in registry
by Nerijus Baliunas Oct. 9, 2001

Oct. 9, 2001

On Mon, 8 Oct 2001 22:36:27 +0200 Stefan Leichter <Stefan.Leichter(a)camline.com> wrote: SL> Hello, SL> SL> during some investigations on CEMW i found a posting from Nerijus Baliunas SL> <nerijus(a)users.sourceforge.net> that contains the registry information to SL> make the installer of IE 4.0 thinking the SP3 is installed on a NT 4.0 system. SL> SL> Tschüß Stefan SL> SL> --- wine-20011004/winedefault.reg Fri Oct 5 13:12:57 2001 SL> +++ wine/winedefault.reg Mon Oct 8 22:25:33 2001 SL> @@ -100,6 +100,13 @@ SL> "OEMCP"="437" SL> SL> # SL> +# Win NT 4.0 with Service Pack 3 SL> +# SL> + SL> +[HKEY_LOCAL_MACHINE\System\CurrentControlSet\Control\Windows] SL> +"CSDVersion"="768" Maybe better to use "1536"? It means Service Pack 6. Regards, Nerijus

1 0

Re: SetSystemMetrics
by Alexandre Julliard Oct. 9, 2001

Oct. 9, 2001

Andriy Palamarchuk <apa3a(a)yahoo.com> writes: > Files sysmetrics.c, sysmetrics.h: > Implementation of SetSystemMetrics function which sets data accessed by > GetSystemMetrics. Applied, but I renamed it to SYSMETRICS_Set. Internal functions should not be named the same way as exported APIs, otherwise people reading the code will think this is a standard Windows function. -- Alexandre Julliard julliard(a)winehq.com

2 1

using a webcam driver for windows
by Hrafnkell Eiríksson Oct. 9, 2001

Oct. 9, 2001

Hi This might be a silly question, but would it be possible to use wine with a parallel port webcam driver and an windows application to show images to use a webcam in linux that only has windows suport? I'm thinking about how to get my Logitech QuickCam VC to work in linux :) I know that wine can to some extent access the parallel port in linux. I'm not very familiar with the inner workings of windows, but I guess the camera driver is something called .VxD? Would it be possible to let wine use that somehow? Or does wine not work at that low level? If it is possible to do this, is there something that a guy armed with a C compiler but almost no understanding of windows and wine internals can get working? Thanks Hrafnkell

1 0

Rendering problem in NetTerm
by Mike McCormack Oct. 9, 2001

Oct. 9, 2001

Hi All, i'm trying to solve a comms problem in NetTerm http://www.securenetterm.com/ but i've run into another problem. The rendering in NetTerm fails almost completely. Its main window ends up with only the boarders drawn and will draw some toolbar buttons if i move the mouse over them. The following message appears repeatedly in the debug log: err:region:CombineRgn Invalid rgn=1096 i did a bit of tracing of the app and found that the region 1096 is created, used by a DC, destroyed, then used again. Is it possible that the reference count of the region is incremented when it is used in some way by a DC? A similar problem to the reference counting one Uwe Bonnes discovered (see message to this list on 16 Aug)? Mike >---------- Forwarded message ---------- >Date: Fri, 14 Sep 2001 18:41:04 -0300 >From: Raul Dias <chaos(a)swi.com.br> >Reply-To: wine-users(a)winehq.com >To: wine-users(a)winehq.com >Subject: wine serial support > >Hi, > >I use wine to run a terminal emulator (NetTerm). >It connects to a Linux Server (it was meant to run from windows) that runs a commerce software. > >This Linux commerce software sends the terminal emulator escape codes to interface with a >"ticket" printer attached to a serial port. > >The reason to go thru all this problem is that the printer has a proprietary, closed source driver >for windows only and interface only with netterm. > >With wine 20001026 this works greatfull. The application takes 3 seconds to startup >in a P133 machine. > >However, the serial communication part does not work with the later versions of wine >like 20010731. The application reports as if there is no serial port and wine complains >with some fixmes about comm support. > >Did wine suffered a big serial change over the last year? > >Now the important thing: >What can I do to help debug this problem (and make it work >on latest wine versions)? >Run it in the latest wine with --debugmsg +all is enough? > > >Regards, > >Raul Dias ------------------------------------------ mailto:Mike_McCormack(a)start.com.au ph +82 16 430 0425 __________________________________________________________________ Get your free Australian email account at http://www.Looksmart.com.au

2 1

Re: CommandLineToArgvW fix
by Alexandre Julliard Oct. 9, 2001

Oct. 9, 2001

Francois Gouget <fgouget(a)free.fr> writes: > The problem is that CommandLineToArgvW is supposed to return a > pointer, not a handle. But I cannot do: > > hargv=GlobalAlloc(...); > argv=GlobalLock(hargv); > ... > return argv; Sure you can do that, as long as the allocation is GMEM_FIXED GlobalLock doesn't touch the lock count so you don't need to unlock it. It's basically just a glorified typecast. > Sigh. What a stupid memory API. Here I have to agree... -- Alexandre Julliard julliard(a)winehq.com

1 0

Re: CommandLineToArgvW fix
by Alexandre Julliard Oct. 9, 2001

Oct. 9, 2001

Francois Gouget <fgouget(a)free.fr> writes: > Grrr, HGLOBAL is not a void* pointer... yet => casts :-( GlobalLock would be better... -- Alexandre Julliard julliard(a)winehq.com

2 1