April 2003 - Wine-Devel - list.winehq.org

Re: REFIXED: [dx8-7] Fix compile problem introduced by dx8-2
by keisangi April 20, 2003

April 20, 2003

Ann and Jason Edmeades wrote: >Please use this fix rather than the last one - Just noticed the break >statement was the wrong side of the #endif > >Changelog > >DOT3 isnt there on earlier opengl versions > >Jason > > hi, i've subscribed this mailling list for testing purposes only, now i'd like to unsubscribe but i can't figure how to do. could someone can tell me?

3 2

Re: Remove db2html-winehq
by Alexandre Julliard April 20, 2003

April 20, 2003

"Dimitrie O. Paun" <dimi(a)intelliware.ca> writes: > There is no reason to have our own db2html hack. It creates > problems, we never got it to work reliably on different distros, > etc. We should use the one provided with the system, it's a far > safer bet. Also, our build system uses the standard db2html, > it's only the make_winehq script that used it. And that script > is used only on WineHQ, where we should be able to control the > environment to have a half-decent db2html that supports the > -d option. Do we actually need that make_winehq script at all? It seems almost everything it does is already done in the normal makefiles, except the bit about the winehq stylesheet. And I'm not sure what this one is for, since the docs on WineHQ look identical to the one we get by doing 'make doc'. Does anyone know how this is all supposed to work? -- Alexandre Julliard julliard(a)winehq.com

3 6

cvs woes
by Ann and Jason Edmeades April 20, 2003

April 20, 2003

| device.c: In function `IDirect3DDevice8Impl_SetTextureStageState': | device.c:3456: `GL_MIRRORED_REPEAT_ARB' undeclared (first use in this function) | device.c:3456: (Each undeclared identifier is reported only once | device.c:3456: for each function it appears in.) Hmm - This is a more tricky one to fix - Lionel, got any ideas? I have #if'defed it for GL_VERSION_1_3, but it looks like some headers who define GL_VERSION_1_3 dont have the mirrored repeat ARB defines. As a workaround, add #define GL_MIRRORED_REPEAT_ARB 0x8370 at the top of the source part. Question - If you look in your glext.h, does it define GL_MIRRORED_REPEAT instead? Jason

3 3

compiling wine under cygwin - status ??
by Kai Kai April 20, 2003

April 20, 2003

Hi, all, I am a NEWBIE in this area. I am very interested in the wine and cygwin. Can wine work under cygwin now? Or can Wine DLL be build under cygwin? I just found the status from David on Nov28, 2002, is there news about it now? howto is better:-) Thanks Kai _________________________________________________________ Do You Yahoo!? 雅虎邀你参与CCTV-2对话节目调查：移动、联通、灵通，要说爱你不容易 http://rd.yahoo.com/mail_cn/tag/?http://cn.surveys.yahoo.com/cctv_wireless_…

1 0

CVS build failed
by Paul Millar April 20, 2003

April 20, 2003

Anyone else getting this? On a fairly standard RH7.3 machine: device.c: In function `IDirect3DDevice8Impl_SetTextureStageState': device.c:3413: `GL_DOT3_RGBA' undeclared (first use in this function) device.c:3413: (Each undeclared identifier is reported only once device.c:3413: for each function it appears in.) make[2]: *** [device.o] Error 1 Looks like http://cvs.winehq.com/patch.py?id=7915 is the cause. See http://www.astro.gla.ac.uk/users/paulm/WRT/wrt.php and http://www.astro.gla.ac.uk/users/paulm/WRT/buildinfo.php?id=394 for more details .. Cheers, Paul. ---- Paul Millar

3 2

Re: UPDATED: video mode changes
by Lionel Ulmer April 20, 2003

April 20, 2003

Just a small nitpick : > + XF86VidModeModeInfo *mode; > + DWORD dwBpp = GetSystemMetrics(SM_WINE_BPP); > + if (dwBpp == 24) dwBpp = 32; I would prefer this change to be done directly in the 'GetDeviceCaps' function of the X11DRV to always return 32 in case the scren depth is 24. I know for sure that this fixes at least one application and should not hurt others :-) A part from that, the patch looks nice... If it breaks apps on Desktop mode, we will fix it then by re-adding the 'dummy' modes :-) Lionel -- Lionel Ulmer - http://www.bbrox.org/

2 2

Re: [dx8-5] Windows never reports 24 bit screen depths
by Ann and Jason Edmeades April 19, 2003

April 19, 2003

(Apologies for the duplicate sent - Didnt put a subject on first time) >> Windows never reports 24 bit colours, and some games fail to work unless in >> specific modes. Report 16bit colour for 24bit modes >Out of curiosity, in that case, why report 24 bits at all ? Why not just >report 8, 16 and 32 ? I dont - I query it with GetDeviceCaps and this is what gets returned. hdc = CreateDCA("DISPLAY", NULL, NULL, NULL); bpp = GetDeviceCaps(hdc, BITSPIXEL); DeleteDC(hdc); Actually, all the screen handling in the dx8 code is awful (I can say that as I put it there...). I am hoping that if the recent changes to the window modes is implemented, I can use that code to query what modes this device supports, and to change the mode/resolution appropriately. The code that we have is a semi working stopgap really. Jason

2 3

(no subject)
by Ann and Jason Edmeades April 19, 2003

April 19, 2003

>> Windows never reports 24 bit colours, and some games fail to work unless in >> specific modes. Report 16bit colour for 24bit modes >Out of curiosity, in that case, why report 24 bits at all ? Why not just >report 8, 16 and 32 ? I dont - I query it with GetDeviceCaps and this is what gets returned. hdc = CreateDCA("DISPLAY", NULL, NULL, NULL); bpp = GetDeviceCaps(hdc, BITSPIXEL); DeleteDC(hdc); Actually, all the screen handling in the dx8 code is awful (I can say that as I put it there...). I am hoping that if the recent changes to the window modes is implemented, I can use that code to query what modes this device supports, and to change the mode/resolution appropriately. The code that we have is a semi working stopgap really. Jason

1 0

Re: [dx8-5] Windows never reports 24 bit screen depths
by Lionel Ulmer April 19, 2003

April 19, 2003

> Windows never reports 24 bit colours, and some games fail to work unless in > specific modes. Report 16bit colour for 24bit modes Out of curiosity, in that case, why report 24 bits at all ? Why not just report 8, 16 and 32 ? Lionel -- Lionel Ulmer - http://www.bbrox.org/

1 0

Re: [Resend] Implements languagegroups functions
by Alexandre Julliard April 19, 2003

April 19, 2003

Maxime Bellengé <maxime.bellenge(a)laposte.net> writes: > This patch implements the languagegroups functions. > As I have access to a French and English windows 2k+, I set > the resources accordingly for the kernel32 resources. However, as > I don't know the resources for other languages, I have put English > resources in the other nls files. People will have to change them to the > correct values. > As suggested by alexandre, I chose some unused resource ids because even > win2k and winxp don't store them the same way and both of them > conflicted with the actual kernel32 resources from wine. I applied the resource strings, but the rest will still need a bit of work: you can't use a global variable to store the callback function, this is not reentrant, you need to do everything on the stack. Also I think the locale->language group mapping should be stored in the resources too, not in the code. And I'd suggest pretending that all language groups are installed, since we can't install them dynamically anyway. -- Alexandre Julliard julliard(a)winehq.com

1 0

winedbg
by Michal Janusz Miroslaw April 19, 2003

April 19, 2003

Hello, Is anyone working on winedbg ? I have some ideas and would like to hear opinions/more ideas/flames: * enabling/disabling of displays [done] * displays shown in one function only [done] * implement define command to add symbols on the fly (didn't find it is sources except in help text ...) * save/restore breakpoints/displays/defined symbols * reverse backtrace list (so current frame is at the bottom and you dont have to scroll all the way up) * remove limit of 10 entries in DEBUG_GetSymbolValue [done] * remove duplicated code in DEBUG_FindNearestSymbol/DEBUG_CheckLinenoStatus I haven't looked everywhere, yet, but `grep FIXME *|wc -l` shows 106 now, so I should find something. :) -- Michal Miroslaw

3 4

Wine keyboard focus and bug 1385
by Roderick Colenbrander April 19, 2003

April 19, 2003

Hi, When playing for example the game Jedi Outcast on Wine (quake3 based) in non-deskop a keyboard related problem appears. The game starts fine but doesn't receive any keyboard input. When I press for example the A key, the character appears in the console. I saw this problem before and it looks a lot like bug 1385: http://bugs.winehq.com/show_bug.cgi?id=1385 After looking a bit at the source code I found a way to fix my problem. I edited some code in the SetFocus function from x11drv/window.c. The original code looks like this: /* we must not use CurrentTime (ICCCM), so try to use last message time instead */ /* FIXME: this is not entirely correct */ XSetInputFocus( display, win, RevertToParent, /*CurrentTime*/ GetMessageTime() + X11DRV_server_startticks ); When I set the time to CurrentTime the problem was gone. According to the description and the FIXME neither solutions are correct. I don't know enough about the working of the x11 driver and X programming, so perhaps someone can help? Thanks, Roderick Colenbrander

1 0

Palettes - Any experts?
by Ann and Jason Edmeades April 19, 2003

April 19, 2003

Hello, Having taken a break for a few months I thought I'd try playing with wine again... I have been trying to work out why one game I am investigating never even shows its splash screen, and have got most of the way until I have got stuck. Can anyone offer me some advice? The game launcher kicks off a process whose first thread creates a dummy 16x16 window which is placed in the center of the screen (who knows why!), and launches another thread. The main thread then proceeds to create another process, and wait for it to complete. This means the dummy window is not processing any messages and if sent any the whole thing hangs :-( The launched thread creates the splash screen and begins a WM_PAINT operation. During this processing it calls CreateCompatibleDC, SelectObject, SelectPalette and finally RealizePalette This RealizePalette call (UserRealizePalette, windows\painting.c) results in 256 colors being mapped in, and SendMessageA( HWND_BROADCAST, WM_PALETTECHANGED, (WPARAM)hWnd, 0L); to occur. This broadcast hangs attempting to send the message to the first thread as it is in a hung state waiting for another process to end. Under windows, the RealizePalette returns 0, and spy on Windows does not show this message being sent to the 'hung' thread. 'Interestingly', putting an xterm behind the splash panel, and trying to bring it to the front (to force a repaint) shows that UserRealizePalette is always returning 256, which I thing is the underlying bug. It appears to create a palette every WM_PAINT operation. Commenting out the broadcast makes the splashscreen appear perfectly, so I know I am along the right lines. Can anyone fill in some gaps and point me in the right direction? Debug logs willingly available! :-) Thanks Jason

3 6

FreeBSD and Total Annihilation (Bug 1392)
by Rod Taylor April 19, 2003

April 19, 2003

With the sound fix, the demo version works great. The real version (which I just bought) doesn't work at all (blank screen) as shipped or with patches from the distributor. Attached is a trace of dplay and virtual, and below is the code snippit which I believe to be causing the problem. FreeBSD 4.8-RELEASE using 20030408 Wine sources ('official port') Any suggestions? Interesting lines from trace: trace:dplay:DPLAYX_ConstructData Through wait trace:dplay:DPLAYX_ConstructData File mapped 0x5c created trace:virtual:NtMapViewOfSection handle=0x5c addr=0x50000000 off=000000000 size=0 access=4 err:dplay:DPLAYX_ConstructData : unable to map static data into process memory space (487) trace:virtual:NtAllocateVirtualMemory 0x0 00110000 2000 00000040 The source of the error (487) seems to be line 466 in ntdll/virtual.c: static NTSTATUS anon_mmap_aligned( void **addr, unsigned int size, int prot, int flags ) { void *ptr, *base = *addr; unsigned int view_size = size + (base ? 0 : granularity_mask + 1); if ((ptr = wine_anon_mmap( base, view_size, prot, flags )) == (void *)-1) { if (errno == ENOMEM) return STATUS_NO_MEMORY; return STATUS_INVALID_PARAMETER; } if (!base) { /* Release the extra memory while keeping the range * starting on the granularity boundary. */ if ((unsigned int)ptr & granularity_mask) { unsigned int extra = granularity_mask + 1 - ((unsigned int)ptr & granularity_mask); munmap( ptr, extra ); ptr = (char *)ptr + extra; view_size -= extra; } if (view_size > size) munmap( (char *)ptr + size, view_size - size ); } else if (ptr != base) { /* We couldn't get the address we wanted */ munmap( ptr, view_size ); return STATUS_CONFLICTING_ADDRESSES; /* !! HERE !! */ } -- Rod Taylor <rbt(a)rbt.ca> PGP Key: http://www.rbt.ca/rbtpub.asc

1 0

named pipe patch
by Dan Kegel April 18, 2003

April 18, 2003

Here's a draft of a patch that tries to fix two problems: 1. can't open a pipe until server has called ConnectNamedPipe (on Win32, pipes can be opened immediately after CreateNamedPipe) 2. client loses any data in pipe if server calls FlushFileBuffers and then immediately calls DisconnectNamedPipe. I work around this by keeping a reference to the server side fd in the client pipe user. FlushFileBuffers still doesn't work, but we don't care; this is kind of a unixification of named pipe semantics. It shifts our win32 noncompliance from a very touchy spot to a less touchy spot, I hope. In both cases, I try to solve the problem by passing fd's inside wineserver. I have no idea if I'm doing this right, but it does pass most of my single-process regression tests. (Overlapped I/O is still broken, and I'm not sure if this fd passing magic will work if the pipe ends are in different processes.) Interestingly, it seems that the DisconnectNamedPipe()/ConnectNamedPipe() sequence as previously written would have failed if a unix file descriptor was created or deleted between pipe disconnection and reconnection. This is because a new fd is used for each pipe session, but the old fd number was never removed from the handle-to-unix-fd cache upon DisconnectNamedPipe. My workaround for the data loss problem exposed this issue. I "fixed" it by invalidating the cached fd for the handle passed to DisconnectNamedPipe. (If there are any other handles to this object, their fds won't be invalidated. Shouldn't matter in practice...) This patch also includes an improved test for named pipes. I'm posting it here rather than to wine-patches to get feedback on whether the fd passing I'm doing is sane. Thanks, Dan -- Dan Kegel http://www.kegel.com http://counter.li.org/cgi-bin/runscript/display-person.cgi?user=78045 Index: server/object.c =================================================================== RCS file: /home/wine/wine/server/object.c,v retrieving revision 1.27 diff -d -u -r1.27 object.c --- server/object.c 19 Feb 2003 00:33:33 -0000 1.27 +++ server/object.c 18 Apr 2003 17:41:17 -0000 @@ -193,6 +193,7 @@ struct object *grab_object( void *ptr ) { struct object *obj = (struct object *)ptr; + assert( obj ); assert( obj->refcount < INT_MAX ); obj->refcount++; return obj; Index: server/file.h =================================================================== RCS file: /home/wine/wine/server/file.h,v retrieving revision 1.7 diff -d -u -r1.7 file.h --- server/file.h 26 Mar 2003 23:41:43 -0000 1.7 +++ server/file.h 18 Apr 2003 17:41:17 -0000 @@ -49,6 +49,7 @@ extern struct fd *create_anonymous_fd( const struct fd_ops *fd_user_ops, int unix_fd, struct object *user ); extern void *get_fd_user( struct fd *fd ); +extern void set_fd_user( struct fd *fd, struct object *user ); extern int get_unix_fd( struct fd *fd ); extern void fd_poll_event( struct fd *fd, int event ); extern int check_fd_events( struct fd *fd, int events ); Index: server/fd.c =================================================================== RCS file: /home/wine/wine/server/fd.c,v retrieving revision 1.8 diff -d -u -r1.8 fd.c --- server/fd.c 26 Mar 2003 01:32:18 -0000 1.8 +++ server/fd.c 18 Apr 2003 17:41:18 -0000 @@ -884,6 +884,12 @@ return fd->user; } +/* change the object that is using an fd */ +void set_fd_user( struct fd *fd, struct object *user ) +{ + fd->user = user; +} + /* retrieve the unix fd for an object */ int get_unix_fd( struct fd *fd ) { @@ -993,15 +999,22 @@ return fd; } -/* flush a file buffers */ +/* flush a file's buffers */ DECL_HANDLER(flush_file) { - struct fd *fd = get_handle_fd_obj( current->process, req->handle, 0 ); - - if (fd) - { - fd->fd_ops->flush( fd ); - release_object( fd ); + struct object *obj = get_handle_obj(current->process, req->handle, 0, NULL); + if (!obj) { + set_error( STATUS_INVALID_HANDLE ); + } else { + struct fd *fd = get_obj_fd(obj); + /* It's ok for named pipes, and possibly other kinds of objects, + * to not have fds associated sometimes, so do nothing if no fd. + */ + if (fd) { + fd->fd_ops->flush( fd ); + release_object( fd ); + } + release_object( obj ); } } Index: server/handle.h =================================================================== RCS file: /home/wine/wine/server/handle.h,v retrieving revision 1.12 diff -d -u -r1.12 handle.h --- server/handle.h 14 Feb 2003 20:27:10 -0000 1.12 +++ server/handle.h 18 Apr 2003 17:41:18 -0000 @@ -39,6 +39,7 @@ extern struct object *get_handle_obj( struct process *process, obj_handle_t handle, unsigned int access, const struct object_ops *ops ); extern int get_handle_unix_fd( struct process *process, obj_handle_t handle, unsigned int access ); +extern void invalidate_handle_unix_fd( struct process *process, obj_handle_t handle ); extern int set_handle_info( struct process *process, obj_handle_t handle, int mask, int flags, int *fd ); extern obj_handle_t duplicate_handle( struct process *src, obj_handle_t src_handle, struct process *dst, unsigned int access, int inherit, int options ); Index: server/handle.c =================================================================== RCS file: /home/wine/wine/server/handle.c,v retrieving revision 1.26 diff -d -u -r1.26 handle.c --- server/handle.c 19 Feb 2003 00:33:33 -0000 1.26 +++ server/handle.c 18 Apr 2003 17:41:19 -0000 @@ -397,6 +397,17 @@ return entry->fd; } +/* invalidate the cached fd for a given handle */ +void invalidate_handle_unix_fd( struct process *process, obj_handle_t handle ) +{ + struct handle_entry *entry; + + entry = get_handle( process, handle ); + /*fprintf(stderr, "invalidate_handle_unix_fd: handle %d, clearing client unix process fd %d\n", (int)handle, entry->fd);*/ + if (entry) + entry->fd = -1; +} + /* find the first inherited handle of the given type */ /* this is needed for window stations and desktops (don't ask...) */ obj_handle_t find_inherited_handle( struct process *process, const struct object_ops *ops ) Index: server/named_pipe.c =================================================================== RCS file: /home/wine/wine/server/named_pipe.c,v retrieving revision 1.22 diff -d -u -r1.22 named_pipe.c --- server/named_pipe.c 4 Apr 2003 22:26:34 -0000 1.22 +++ server/named_pipe.c 18 Apr 2003 17:41:19 -0000 @@ -48,12 +48,13 @@ enum pipe_state { ps_none, - ps_idle_server, - ps_wait_open, - ps_wait_connect, + ps_fresh_server, /* server after create, before first connect */ + ps_idle_server, /* server after disconnect */ + ps_wait_open, /* server during ConnectNamedPipe() wait */ + ps_wait_connect, /* client during WaitNamedPipe() */ ps_connected_server, ps_connected_client, - ps_disconnected + ps_disconnected, /* client after disconnect */ }; struct named_pipe; @@ -62,6 +63,7 @@ { struct object obj; struct fd *fd; + struct fd *fd_for_peer; enum pipe_state state; struct pipe_user *other; struct named_pipe *pipe; @@ -103,6 +105,7 @@ static void pipe_user_destroy( struct object *obj); static int pipe_user_get_poll_events( struct fd *fd ); +static int pipe_user_flush( struct fd *fd ); static int pipe_user_get_info( struct fd *fd, struct get_file_info_reply *reply, int *flags ); static const struct object_ops pipe_user_ops = @@ -121,7 +124,7 @@ { pipe_user_get_poll_events, /* get_poll_events */ default_poll_event, /* poll_event */ - no_flush, /* flush */ + pipe_user_flush, /* flush */ pipe_user_get_info, /* get_file_info */ no_queue_async /* queue_async */ }; @@ -163,6 +166,8 @@ static struct fd *pipe_user_get_fd( struct object *obj ) { struct pipe_user *user = (struct pipe_user *)obj; + if (!user->fd) + return NULL; return (struct fd *)grab_object( user->fd ); } @@ -177,15 +182,20 @@ if(user->other) { - release_object( user->other->fd ); - user->other->fd = NULL; switch(user->other->state) { case ps_connected_server: user->other->state = ps_idle_server; + release_object( user->other->fd ); + user->other->fd = NULL; break; case ps_connected_client: user->other->state = ps_disconnected; + if (user->fd) { + user->other->fd_for_peer = user->fd; + set_fd_user(user->fd, &user->other->obj); + user->fd = NULL; + } break; default: fprintf(stderr,"connected pipe has strange state %d!\n", @@ -202,6 +212,7 @@ if (user->thread) release_object(user->thread); release_object(user->pipe); if (user->fd) release_object( user->fd ); + if (user->fd_for_peer) release_object(user->fd_for_peer); } static int pipe_user_get_poll_events( struct fd *fd ) @@ -209,6 +220,14 @@ return POLLIN | POLLOUT; /* FIXME */ } +static int pipe_user_flush( struct fd *fd ) +{ + /* Pretend we work. We don't, but the fact that we don't + * destroy contents on disconnect makes up for it in most cases. + */ + return 0; +} + static int pipe_user_get_info( struct fd *fd, struct get_file_info_reply *reply, int *flags ) { if (reply) @@ -273,6 +292,7 @@ return NULL; user->fd = NULL; + user->fd_for_peer = NULL; user->pipe = pipe; user->state = ps_none; user->other = NULL; @@ -290,13 +310,14 @@ return user; } -static struct pipe_user *find_partner(struct named_pipe *pipe, enum pipe_state state) +/* Find a pipe_user in any of the given states */ +static struct pipe_user *_find_partner(struct named_pipe *pipe, int states) { struct pipe_user *x; for(x = pipe->users; x; x=x->next) { - if(x->state==state) + if((1 << x->state) & states) break; } @@ -306,6 +327,12 @@ return (struct pipe_user *)grab_object( x ); } +/* Find a pipe_user in the given state */ +static inline struct pipe_user *find_partner(struct named_pipe *pipe, enum pipe_state s) +{ + return _find_partner(pipe, 1<<s); +} + DECL_HANDLER(create_named_pipe) { struct named_pipe *pipe; @@ -329,7 +356,14 @@ if(user) { - user->state = ps_idle_server; + int fds[2]; + user->state = ps_fresh_server; + if (!socketpair(PF_UNIX, SOCK_STREAM, 0, fds)) + { + user->fd = create_anonymous_fd( &pipe_user_fd_ops, fds[1], &user->obj ); + user->fd_for_peer = create_anonymous_fd( &pipe_user_fd_ops, fds[0], &user->obj ); + /* fprintf(stderr, "create_named_pipe: user %p, created fd %p, created fd_for_peer %p\n", user, user->fd, user->fd_for_peer); */ + } reply->handle = alloc_handle( current->process, user, GENERIC_READ|GENERIC_WRITE, 0 ); release_object( user ); } @@ -349,7 +383,7 @@ set_error( STATUS_NO_SUCH_FILE ); return; } - if (!(partner = find_partner(pipe, ps_wait_open))) + if (!(partner = _find_partner(pipe, (1<<ps_wait_open) | (1<<ps_fresh_server)))) { release_object(pipe); set_error( STATUS_PIPE_NOT_AVAILABLE ); @@ -359,22 +393,36 @@ { int fds[2]; - if(!socketpair(PF_UNIX, SOCK_STREAM, 0, fds)) - { - user->fd = create_anonymous_fd( &pipe_user_fd_ops, fds[1], &user->obj ); - partner->fd = create_anonymous_fd( &pipe_user_fd_ops, fds[0], &partner->obj ); - if (user->fd && partner->fd) - { - notify_waiter(partner,STATUS_SUCCESS); - partner->state = ps_connected_server; - partner->other = user; - user->state = ps_connected_client; - user->other = partner; - reply->handle = alloc_handle( current->process, user, req->access, 0 ); - } - } - else file_set_error(); + if (partner->fd_for_peer != NULL) { + /* Server side is fresh, so it already created the fds. + * Grab the fd it created for us, and set its user (since + * it couldn't know ahead of time). + */ + user->fd = partner->fd_for_peer; + partner->fd_for_peer = NULL; + set_fd_user(user->fd, &user->obj); + /* fprintf(stderr, "open_named_pipe: user %p grabbing fd %p from partner %p\n", user, user->fd, partner); */ + } else if (!socketpair(PF_UNIX, SOCK_STREAM, 0, fds)) { + /* Server side isn't fresh, so we have to create the fds for + * both sides. + */ + user->fd = create_anonymous_fd( &pipe_user_fd_ops, fds[1], &user->obj ); + partner->fd = create_anonymous_fd( &pipe_user_fd_ops, fds[0], &partner->obj ); + /* fprintf(stderr, "open_named_pipe: user %p, created fd %p, created fd %p for partner %p\n", user, user->fd, partner->fd, partner); */ + } else + file_set_error(); + if (user->fd && partner->fd) + { + enum pipe_state old_partner_state = partner->state; + partner->state = ps_connected_server; + partner->other = user; + user->state = ps_connected_client; + user->other = partner; + reply->handle = alloc_handle( current->process, user, req->access, 0 ); + if (old_partner_state == ps_wait_open) + notify_waiter(partner,STATUS_SUCCESS); + } release_object( user ); } release_object( partner ); @@ -389,7 +437,7 @@ if(!user) return; - if( user->state != ps_idle_server ) + if( (user->state & (ps_idle_server|ps_fresh_server)) == 0 ) { set_error(STATUS_PORT_ALREADY_SET); } @@ -455,16 +503,28 @@ if( (user->state == ps_connected_server) && (user->other->state == ps_connected_client) ) { - release_object( user->other->fd ); + /* This duplicates a fair bit of code from the destroy method */ + if (user->fd) { + /* fprintf(stderr, "disconnect_named_pipe: user %p passing fd %p to partner %p\n", user, user->fd, user->other); */ + user->other->fd_for_peer = user->fd; + set_fd_user(user->fd, &user->other->obj); + user->fd = NULL; + } user->other->fd = NULL; user->other->state = ps_disconnected; user->other->other = NULL; - - release_object( user->fd ); - user->fd = NULL; - user->state = ps_idle_server; user->other = NULL; + user->state = ps_idle_server; } + + /* IMPOSSIBLE PART: invalidate the cached unix fd for all + * handles that refer to user. + * We can only do it for the handle we were passed, + * so this will only work if that handle has no duplicate. + * FIXME: find out what happens to a handle in win32 when you + * disconnect a duplicate of it. + */ + invalidate_handle_unix_fd(current->process, req->handle); release_object(user); } Index: dlls/kernel/tests/pipe.c =================================================================== RCS file: /home/wine/wine/dlls/kernel/tests/pipe.c,v retrieving revision 1.2 diff -d -u -r1.2 pipe.c --- dlls/kernel/tests/pipe.c 25 Feb 2003 03:56:43 -0000 1.2 +++ dlls/kernel/tests/pipe.c 18 Apr 2003 17:41:20 -0000 @@ -21,23 +21,40 @@ #include <stdlib.h> #include <stdio.h> #include <time.h> +#include <assert.h> + +#include <wtypes.h> +#include <windef.h> +#include <winbase.h> +#include <winerror.h> #ifndef STANDALONE #include "wine/test.h" +#define trace2(a,b) trace((a),(b)) #else #include <assert.h> #define START_TEST(name) main(int argc, char **argv) #define ok(condition, msg) assert(condition) -#define todo_wine +#define todo_wine if (0) +static void trace2(const char *s, const char *t) +{ + DWORD cbWritten; + char buf[512]; + sprintf(buf, s, t); + WriteFile(GetStdHandle(STD_ERROR_HANDLE), buf, strlen(buf), &cbWritten, NULL); +} +static void trace(const char *s) +{ + DWORD cbWritten; + char buf[512]; + sprintf(buf, "%s", s); + WriteFile(GetStdHandle(STD_ERROR_HANDLE), buf, strlen(buf), &cbWritten, NULL); +} #endif -#include <wtypes.h> -#include <windef.h> -#include <winbase.h> -#include <winerror.h> - #define PIPENAME "\\\\.\\PiPe\\tests_" __FILE__ + void test_CreateNamedPipeA(void) { HANDLE hnp; @@ -45,8 +62,9 @@ const char obuf[] = "Bit Bucket"; char ibuf[32]; DWORD written; - DWORD gelesen; + DWORD nbytes_read; + trace("test_CreateNamedPipe starting\n"); /* Bad parameter checks */ hnp = CreateNamedPipeA("not a named pipe", PIPE_ACCESS_DUPLEX, PIPE_TYPE_BYTE|PIPE_WAIT, @@ -86,10 +104,8 @@ /* lpSecurityAttrib */ NULL); ok(hnp != INVALID_HANDLE_VALUE, "CreateNamedPipe failed"); - hFile = CreateFileA(PIPENAME, GENERIC_READ|GENERIC_WRITE, 0, - NULL, OPEN_EXISTING, 0, 0); - todo_wine - { + hFile = CreateFileA(PIPENAME, GENERIC_READ | GENERIC_WRITE, 0, NULL, OPEN_EXISTING, 0, 0); + todo_wine { ok(hFile != INVALID_HANDLE_VALUE, "CreateFile failed"); } @@ -101,15 +117,15 @@ memset(ibuf, 0, sizeof(ibuf)); ok(WriteFile(hnp, obuf, sizeof(obuf), &written, NULL), "WriteFile"); ok(written == sizeof(obuf), "write file len"); - ok(ReadFile(hFile, ibuf, sizeof(obuf), &gelesen, NULL), "ReadFile"); - ok(gelesen == sizeof(obuf), "read file len"); + ok(ReadFile(hFile, ibuf, sizeof(obuf), &nbytes_read, NULL), "ReadFile"); + ok(nbytes_read == sizeof(obuf), "read file len"); ok(memcmp(obuf, ibuf, written) == 0, "content check"); memset(ibuf, 0, sizeof(ibuf)); ok(WriteFile(hFile, obuf, sizeof(obuf), &written, NULL), "WriteFile"); ok(written == sizeof(obuf), "write file len"); - ok(ReadFile(hnp, ibuf, sizeof(obuf), &gelesen, NULL), "ReadFile"); - ok(gelesen == sizeof(obuf), "read file len"); + ok(ReadFile(hnp, ibuf, sizeof(obuf), &nbytes_read, NULL), "ReadFile"); + ok(nbytes_read == sizeof(obuf), "read file len"); ok(memcmp(obuf, ibuf, written) == 0, "content check"); /* Picky conformance tests */ @@ -148,9 +164,460 @@ } ok(CloseHandle(hnp), "CloseHandle"); + + trace("test_CreateNamedPipe returning\n"); +} + +void test_CreateNamedPipe_instances_must_match(void) +{ + HANDLE hnp, hnp2; + + /* Check no mismatch */ + hnp = CreateNamedPipeA(PIPENAME, PIPE_ACCESS_DUPLEX, PIPE_TYPE_BYTE | PIPE_WAIT, + /* nMaxInstances */ 2, + /* nOutBufSize */ 1024, + /* nInBufSize */ 1024, + /* nDefaultWait */ NMPWAIT_USE_DEFAULT_WAIT, + /* lpSecurityAttrib */ NULL); + ok(hnp != INVALID_HANDLE_VALUE, "CreateNamedPipe failed"); + + hnp2 = CreateNamedPipeA(PIPENAME, PIPE_ACCESS_DUPLEX, PIPE_TYPE_BYTE | PIPE_WAIT, + /* nMaxInstances */ 2, + /* nOutBufSize */ 1024, + /* nInBufSize */ 1024, + /* nDefaultWait */ NMPWAIT_USE_DEFAULT_WAIT, + /* lpSecurityAttrib */ NULL); + ok(hnp2 != INVALID_HANDLE_VALUE, "CreateNamedPipe failed"); + + ok(CloseHandle(hnp), "CloseHandle"); + ok(CloseHandle(hnp2), "CloseHandle"); + + /* Check nMaxInstances */ + hnp = CreateNamedPipeA(PIPENAME, PIPE_ACCESS_DUPLEX, PIPE_TYPE_BYTE | PIPE_WAIT, + /* nMaxInstances */ 1, + /* nOutBufSize */ 1024, + /* nInBufSize */ 1024, + /* nDefaultWait */ NMPWAIT_USE_DEFAULT_WAIT, + /* lpSecurityAttrib */ NULL); + ok(hnp != INVALID_HANDLE_VALUE, "CreateNamedPipe failed"); + + todo_wine { + hnp2 = CreateNamedPipeA(PIPENAME, PIPE_ACCESS_DUPLEX, PIPE_TYPE_BYTE | PIPE_WAIT, + /* nMaxInstances */ 1, + /* nOutBufSize */ 1024, + /* nInBufSize */ 1024, + /* nDefaultWait */ NMPWAIT_USE_DEFAULT_WAIT, + /* lpSecurityAttrib */ NULL); + ok(hnp2 == INVALID_HANDLE_VALUE + && GetLastError() == ERROR_PIPE_BUSY, "nMaxInstances not obeyed"); + } + + ok(CloseHandle(hnp), "CloseHandle"); + + /* Check PIPE_ACCESS_* */ + hnp = CreateNamedPipeA(PIPENAME, PIPE_ACCESS_DUPLEX, PIPE_TYPE_BYTE | PIPE_WAIT, + /* nMaxInstances */ 2, + /* nOutBufSize */ 1024, + /* nInBufSize */ 1024, + /* nDefaultWait */ NMPWAIT_USE_DEFAULT_WAIT, + /* lpSecurityAttrib */ NULL); + ok(hnp != INVALID_HANDLE_VALUE, "CreateNamedPipe failed"); + + todo_wine { + hnp2 = CreateNamedPipeA(PIPENAME, PIPE_ACCESS_INBOUND, PIPE_TYPE_BYTE | PIPE_WAIT, + /* nMaxInstances */ 1, + /* nOutBufSize */ 1024, + /* nInBufSize */ 1024, + /* nDefaultWait */ NMPWAIT_USE_DEFAULT_WAIT, + /* lpSecurityAttrib */ NULL); + ok(hnp2 == INVALID_HANDLE_VALUE + && GetLastError() == ERROR_ACCESS_DENIED, "PIPE_ACCESS_* mismatch allowed"); + } + + /* etc, etc */ +} + +/** implementation of alarm() */ +static DWORD CALLBACK alarmThreadMain(LPVOID arg) +{ + DWORD timeout = (DWORD) arg; + trace("alarmThreadMain\n"); + Sleep(timeout); + ok(FALSE, "alarm"); + ExitProcess(1); + return 1; +} + +HANDLE hnp = INVALID_HANDLE_VALUE; + +/** Trivial byte echo server - disconnects after each session */ +static DWORD CALLBACK serverThreadMain1(LPVOID arg) +{ + int i; + + trace("serverThreadMain1 start\n"); + /* Set up a simple echo server */ + hnp = CreateNamedPipeA(PIPENAME "serverThreadMain1", PIPE_ACCESS_DUPLEX, + PIPE_TYPE_BYTE | PIPE_WAIT, + /* nMaxInstances */ 1, + /* nOutBufSize */ 1024, + /* nInBufSize */ 1024, + /* nDefaultWait */ NMPWAIT_USE_DEFAULT_WAIT, + /* lpSecurityAttrib */ NULL); + + ok(hnp != INVALID_HANDLE_VALUE, "CreateNamedPipe failed"); + for (i = 0; ; i++) { + char buf[512]; + DWORD written; + DWORD nbytes_read; + DWORD success; + + /* Wait for client to connect */ + trace("Server calling ConnectNamedPipe...\n"); + ok(ConnectNamedPipe(hnp, NULL) + || GetLastError() == ERROR_PIPE_CONNECTED, "ConnectNamedPipe"); + trace("ConnectNamedPipe returned.\n"); + + /* Echo bytes once */ + memset(buf, 0, sizeof(buf)); + + trace("Server reading...\n"); + success = ReadFile(hnp, buf, sizeof(buf), &nbytes_read, NULL); + trace("Server done reading.\n"); + ok(success, "ReadFile"); + trace2("Server read '%s'\n", buf); + ok(memcmp(buf, "client->server", 14) == 0, "content check"); + + trace("Server writing...\n"); + sprintf(buf, "server->client i %d ", i); + ok(WriteFile(hnp, buf, nbytes_read, &written, NULL), "WriteFile"); + trace("Server done writing.\n"); + ok(written == nbytes_read, "write file len"); + + /* finish this connection, wait for next one */ + trace("Server flushing.\n"); + todo_wine { + ok(FlushFileBuffers(hnp), "FlushFileBuffers"); + } + trace("Server disconnecting.\n"); + ok(DisconnectNamedPipe(hnp), "DisconnectNamedPipe"); + } +} + +/** Trivial byte echo server - closes after each connection */ +static DWORD CALLBACK serverThreadMain2(LPVOID arg) +{ + int i; + HANDLE hnpNext = 0; + + trace("serverThreadMain2\n"); + /* Set up a simple echo server */ + hnp = CreateNamedPipeA(PIPENAME "serverThreadMain2", PIPE_ACCESS_DUPLEX, + PIPE_TYPE_BYTE | PIPE_WAIT, + /* nMaxInstances */ 2, + /* nOutBufSize */ 1024, + /* nInBufSize */ 1024, + /* nDefaultWait */ NMPWAIT_USE_DEFAULT_WAIT, + /* lpSecurityAttrib */ NULL); + ok(hnp != INVALID_HANDLE_VALUE, "CreateNamedPipe failed"); + + for (i = 0; ; i++) { + char buf[512]; + DWORD written; + DWORD nbytes_read; + DWORD success; + + /* Wait for client to connect */ + trace("Server calling ConnectNamedPipe...\n"); + ok(ConnectNamedPipe(hnp, NULL) + || GetLastError() == ERROR_PIPE_CONNECTED, "ConnectNamedPipe"); + trace("ConnectNamedPipe returned.\n"); + + /* Echo bytes once */ + memset(buf, 0, sizeof(buf)); + + trace("Server reading...\n"); + success = ReadFile(hnp, buf, sizeof(buf), &nbytes_read, NULL); + trace("Server done reading.\n"); + ok(success, "ReadFile"); + + trace("Server writing...\n"); + ok(WriteFile(hnp, buf, nbytes_read, &written, NULL), "WriteFile"); + trace("Server done writing.\n"); + ok(written == nbytes_read, "write file len"); + + /* finish this connection, wait for next one */ + trace("Server flushing.\n"); + todo_wine { ok( FlushFileBuffers(hnp), "FlushFileBuffers"); } + trace("Server disconnecting.\n"); + ok(DisconnectNamedPipe(hnp), "DisconnectNamedPipe"); + + /* Set up next echo server */ + hnpNext = + CreateNamedPipeA(PIPENAME "serverThreadMain2", PIPE_ACCESS_DUPLEX, + PIPE_TYPE_BYTE | PIPE_WAIT, + /* nMaxInstances */ 2, + /* nOutBufSize */ 1024, + /* nInBufSize */ 1024, + /* nDefaultWait */ NMPWAIT_USE_DEFAULT_WAIT, + /* lpSecurityAttrib */ NULL); + + ok(hnpNext != INVALID_HANDLE_VALUE, "CreateNamedPipe failed"); + + ok(CloseHandle(hnp), "CloseHandle"); + hnp = hnpNext; + } +} + +/** Trivial byte echo server - uses overlapped named pipe calls */ +static DWORD CALLBACK serverThreadMain3(LPVOID arg) +{ + int i; + HANDLE hEvent; + + trace("serverThreadMain3\n"); + /* Set up a simple echo server */ + hnp = CreateNamedPipeA(PIPENAME "serverThreadMain3", PIPE_ACCESS_DUPLEX | FILE_FLAG_OVERLAPPED, + PIPE_TYPE_BYTE | PIPE_WAIT, + /* nMaxInstances */ 1, + /* nOutBufSize */ 1024, + /* nInBufSize */ 1024, + /* nDefaultWait */ NMPWAIT_USE_DEFAULT_WAIT, + /* lpSecurityAttrib */ NULL); + ok(hnp != INVALID_HANDLE_VALUE, "CreateNamedPipe failed"); + + hEvent = CreateEvent(NULL, // security attribute + TRUE, // manual reset event + FALSE, // initial state + NULL); // name + ok(hEvent != NULL, "CreateEvent"); + + for (i = 0; ; i++) { + char buf[512]; + DWORD written; + DWORD nbytes_read; + DWORD dummy; + DWORD success; + OVERLAPPED oOverlap; + int letWFSOEwait = (i & 2); + int letGORwait = (i & 1); + DWORD err; + + memset(&oOverlap, 0, sizeof(oOverlap)); + oOverlap.hEvent = hEvent; + + /* Wait for client to connect */ + trace("Server calling overlapped ConnectNamedPipe...\n"); + success = ConnectNamedPipe(hnp, &oOverlap); + err = GetLastError(); + ok(success || err == ERROR_IO_PENDING + || err == ERROR_PIPE_CONNECTED, "overlapped ConnectNamedPipe"); + trace("overlapped ConnectNamedPipe returned.\n"); + if (!success && (err == ERROR_IO_PENDING) && letWFSOEwait) + ok(WaitForSingleObjectEx(hEvent, INFINITE, TRUE) == 0, "wait ConnectNamedPipe"); + success = GetOverlappedResult(hnp, &oOverlap, &dummy, letGORwait); + if (!letGORwait && !letWFSOEwait && !success) { + ok(GetLastError() == ERROR_IO_INCOMPLETE, "GetOverlappedResult"); + success = GetOverlappedResult(hnp, &oOverlap, &dummy, TRUE); + } + ok(success, "GetOverlappedResult ConnectNamedPipe"); + trace("overlapped ConnectNamedPipe operation complete.\n"); + + /* Echo bytes once */ + memset(buf, 0, sizeof(buf)); + + trace("Server reading...\n"); + success = ReadFile(hnp, buf, sizeof(buf), NULL, &oOverlap); + trace("Server ReadFile returned...\n"); + err = GetLastError(); + ok(success || err == ERROR_IO_PENDING, "overlapped ReadFile"); + trace("overlapped ReadFile returned.\n"); + if (!success && (err == ERROR_IO_PENDING) && letWFSOEwait) + ok(WaitForSingleObjectEx(hEvent, INFINITE, TRUE) == 0, "wait ReadFile"); + success = GetOverlappedResult(hnp, &oOverlap, &nbytes_read, letGORwait); + if (!letGORwait && !letWFSOEwait && !success) { + ok(GetLastError() == ERROR_IO_INCOMPLETE, "GetOverlappedResult"); + success = GetOverlappedResult(hnp, &oOverlap, &nbytes_read, TRUE); + } + trace("Server done reading.\n"); + ok(success, "overlapped ReadFile"); + + trace("Server writing...\n"); + success = WriteFile(hnp, buf, nbytes_read, NULL, &oOverlap); + trace("Server WriteFile returned...\n"); + err = GetLastError(); + ok(success || err == ERROR_IO_PENDING, "overlapped WriteFile"); + trace("overlapped WriteFile returned.\n"); + if (!success && (err == ERROR_IO_PENDING) && letWFSOEwait) + ok(WaitForSingleObjectEx(hEvent, INFINITE, TRUE) == 0, "wait WriteFile"); + success = GetOverlappedResult(hnp, &oOverlap, &written, letGORwait); + if (!letGORwait && !letWFSOEwait && !success) { + ok(GetLastError() == ERROR_IO_INCOMPLETE, "GetOverlappedResult"); + success = GetOverlappedResult(hnp, &oOverlap, &written, TRUE); + } + trace("Server done writing.\n"); + ok(success, "overlapped WriteFile"); + ok(written == nbytes_read, "write file len"); + + /* finish this connection, wait for next one */ + trace("Server flushing.\n"); + todo_wine { ok( FlushFileBuffers(hnp), "FlushFileBuffers"); } + trace("Server disconnecting.\n"); + ok(DisconnectNamedPipe(hnp), "DisconnectNamedPipe"); + } +} + +static void exercizeServer(const char *pipename, HANDLE serverThread) +{ + int i; + DWORD success; + + trace("exercizeServer starting\n"); + for (i = 0; i < 8; i++) { + HANDLE hFile; + char obuf[32]; + char ibuf[32]; + DWORD written; + DWORD nbytes_read; + int loop; + + for (loop = 0; loop < 3; loop++) { + DWORD err; + trace("Client connecting...\n"); + /* Connect to the server */ + hFile = CreateFileA(pipename, GENERIC_READ | GENERIC_WRITE, 0, + NULL, OPEN_EXISTING, 0, 0); + if (hFile != INVALID_HANDLE_VALUE) + break; + err = GetLastError(); + if (loop == 0) + ok(err == ERROR_PIPE_BUSY || err == ERROR_FILE_NOT_FOUND, "connecting to pipe"); + else + ok(err == ERROR_PIPE_BUSY, "connecting to pipe"); + trace("connect failed, retrying\n"); + Sleep(200); + } + ok(hFile != INVALID_HANDLE_VALUE, "client opening named pipe"); + + /* Make sure it can echo */ + memset(ibuf, 0, sizeof(ibuf)); + sprintf(obuf, "client->server i %d ", i); + trace2("Client writing '%s'...\n", obuf); + ok(WriteFile(hFile, obuf, sizeof(obuf), &written, NULL), "WriteFile to client end of pipe"); + ok(written == sizeof(obuf), "write file len"); + trace("Client reading...\n"); + success = ReadFile(hFile, ibuf, sizeof(ibuf), &nbytes_read, NULL); + if (!success) + printf("ReadFile(%p, %p, %d,,0) failed, err %ld\n", hFile, ibuf, sizeof(ibuf), GetLastError()); + trace2("Client read '%s'\n", ibuf); + ok(success, "ReadFile from client end of pipe"); + ok(nbytes_read == sizeof(obuf), "read file len"); + /*ok(memcmp(obuf, ibuf, written) == 0, "content check");*/ + + trace("Client closing...\n"); + ok(CloseHandle(hFile), "CloseHandle"); + } + + ok(TerminateThread(serverThread, 0), "TerminateThread"); + CloseHandle(hnp); + trace("exercizeServer returning\n"); +} + +void test_NamedPipe_2(void) +{ + HANDLE serverThread; + DWORD serverThreadId; + HANDLE alarmThread; + DWORD alarmThreadId; + + trace("test_NamedPipe_2 starting\n"); + /* Set up a ten second timeout */ + alarmThread = CreateThread(NULL, 0, alarmThreadMain, (void *) 10000, 0, &alarmThreadId); + + /* The servers we're about to exercize do try to clean up carefully, + * but to reduce the change of a test failure due to a pipe handle + * leak in the test code, we'll use a different pipe name for each server. + */ + + trace("Try server #1\n"); + serverThread = CreateThread(NULL, 0, serverThreadMain1, 0, 0, &serverThreadId); + ok(serverThread != INVALID_HANDLE_VALUE, "CreateThread"); + exercizeServer(PIPENAME "serverThreadMain1", serverThread); + + trace("Try server #2\n"); + serverThread = CreateThread(NULL, 0, serverThreadMain2, 0, 0, &serverThreadId); + ok(serverThread != INVALID_HANDLE_VALUE, "CreateThread"); + exercizeServer(PIPENAME "serverThreadMain2", serverThread); + + todo_wine { + trace("Try server #3\n"); + serverThread = CreateThread(NULL, 0, serverThreadMain3, 0, 0, &serverThreadId); + ok(serverThread != INVALID_HANDLE_VALUE, "CreateThread"); + exercizeServer(PIPENAME "serverThreadMain3", serverThread); + + ok(TerminateThread(alarmThread, 0), "TerminateThread"); + trace("test_NamedPipe_2 returning\n"); + } +} + +void test_DisconnectNamedPipe(void) +{ + HANDLE hnp; + HANDLE hFile; + const char obuf[] = "Bit Bucket"; + char ibuf[32]; + DWORD written; + DWORD nbytes_read; + + hnp = CreateNamedPipeA(PIPENAME, PIPE_ACCESS_DUPLEX, PIPE_TYPE_BYTE | PIPE_WAIT, + /* nMaxInstances */ 1, + /* nOutBufSize */ 1024, + /* nInBufSize */ 1024, + /* nDefaultWait */ NMPWAIT_USE_DEFAULT_WAIT, + /* lpSecurityAttrib */ NULL); + ok(hnp != INVALID_HANDLE_VALUE, "CreateNamedPipe failed"); + + ok(WriteFile(hnp, obuf, sizeof(obuf), &written, NULL) == 0 + && GetLastError() == ERROR_PIPE_LISTENING, "WriteFile to not-yet-connected pipe"); + ok(ReadFile(hnp, ibuf, sizeof(ibuf), &nbytes_read, NULL) == 0 + && GetLastError() == ERROR_PIPE_LISTENING, "ReadFile from not-yet-connected pipe"); + + hFile = CreateFileA(PIPENAME, GENERIC_READ | GENERIC_WRITE, 0, NULL, OPEN_EXISTING, 0, 0); + todo_wine { + ok(hFile != INVALID_HANDLE_VALUE, "CreateFile failed"); + } + + /* don't try to do i/o if one side couldn't be opened, as it hangs */ + if (hFile != INVALID_HANDLE_VALUE) { + + /* see what happens if server calls DisconnectNamedPipe + * when there are bytes in the pipe + */ + + ok(WriteFile(hFile, obuf, sizeof(obuf), &written, NULL), "WriteFile"); + ok(written == sizeof(obuf), "write file len"); + ok(DisconnectNamedPipe(hnp), "DisconnectNamedPipe while messages waiting"); + ok(WriteFile(hFile, obuf, sizeof(obuf), &written, NULL) == 0 + && GetLastError() == ERROR_PIPE_NOT_CONNECTED, "WriteFile to disconnected pipe"); + ok(ReadFile(hnp, ibuf, sizeof(ibuf), &nbytes_read, NULL) == 0 + && GetLastError() == ERROR_PIPE_NOT_CONNECTED, + "ReadFile from disconnected pipe with bytes waiting"); + ok(CloseHandle(hFile), "CloseHandle"); + } + + ok(CloseHandle(hnp), "CloseHandle"); + } START_TEST(pipe) { + trace("test 1 of 4:\n"); test_CreateNamedPipeA(); + trace("test 2 of 4:\n"); + test_CreateNamedPipe_instances_must_match(); + trace("test 3 of 4:\n"); + test_NamedPipe_2(); + trace("test 4 of 4:\n"); + test_DisconnectNamedPipe(); + trace("all tests done\n"); }

1 0

winelib on sparc?
by Robert Reif April 18, 2003

April 18, 2003

I would like to do some winelib programming on a sparc and I'm struggling with a few concepts. winelib and wine are the same sources? When you build and run the tests in wine, are the tests really winelib apps? Has anyone really done this? The docs just say "uncharted territory" and don't go any further. I added code for my architecture and compile the wine source but get assembler errors on the spec files. Any help would be welcomed. Thanks, Bob.

1 0

Bug Hunting - uninstaller.exe on Win2K with uninstall information.
by Steven Edwards April 18, 2003

April 18, 2003

Hello Everyone, This bug is a year+ old and I have not done much work on it untill now as I never could debug the stupid thing. This has always bothered me because this program should work fine under Win2K if we are compatible with the registry APIs and such. Does anyone have a idea as to what is going on? Here is the info from bug 1395 in bugzilla When running uninstaller built from Mingw or MS-VC++ 6 under Win2K the application seems to work and run fine when there is no information in the uninstall branch of the registry. As soon as a program is registered and uninstaller is run the program crashes. This has been tested on 3 systems and with many differnt sources and versions of Mingw. When debugging under MSVC with uninstall information in the registry the debugger stops at this code. FetchUninstallInformation() Line 225/226: entries[numentries-1].key = HeapAlloc(GetProcessHeap(), 0, strlen(subKeyName)+1); Entries : Cxx0030: Expression could not be evaluated I will attach a Visual Stdio build with sources/debugging information this evening. Thanks Steven

1 0

Re: remove server silliniess
by Bill Medland April 18, 2003

April 18, 2003

On April 16, 2003 01:51 pm, Gregory M. Turner wrote: > remove run-time checks of compile-time constants. Why is that a silliness? I dearly wish that there were a way of doing compile-time assertion, but there isn't, so it has to be done by a runtime assertion. So it adds a tiny amount to the runtime. If that really bugs you then create a function of compile-time assertions and ensure that it gets called once only. Personally considering the current status of wine and the way we have to work on code that we don't really understand I am surprised at how few assertions there are (and lets not get started on internal code documentation) Or are you saying that assertions are a waste of time :-) -- Bill Medland ACCPAC International, Inc. medbi01(a)accpac.com Corporate: www.accpac.com Hosted Services: www.accpaconline.com

5 5

Re: docu: remove uber document
by Francois Gouget April 18, 2003

April 18, 2003

On Thu, 17 Apr 2003, Dimitrie O. Paun wrote: [...] > Index: documentation/architecture.sgml > =================================================================== > RCS file: /home/wine/wine/documentation/architecture.sgml,v > retrieving revision 1.6 > diff -u -r1.6 architecture.sgml > --- documentation/architecture.sgml 1 Apr 2003 03:26:13 -0000 1.6 > +++ documentation/architecture.sgml 17 Apr 2003 22:09:52 -0000 > @@ -1029,6 +1029,5 @@ >  IMHO, rather than removing these lines, it would be better to modify them to attach the file to the proper book: wine-user.sgml or winelib-user.sgml. -- Francois Gouget fgouget(a)free.fr http://fgouget.free.fr/ f u kn rd ts, ur wy 2 gky 4 ur wn gd.

2 2

Re: rpc server patch
by Gregory M. Turner April 18, 2003

April 18, 2003

On Thursday 17 April 2003 09:36 pm, Ove Kaaven wrote: > diff -u -r1.12 rpc_server.c > --- dlls/rpcrt4/rpc_server.c 18 Apr 2003 02:04:17 -0000 1.12 > +++ dlls/rpcrt4/rpc_server.c 18 Apr 2003 02:16:51 -0000 > @@ -285,11 +286,18 @@ > > static void RPCRT4_new_client(RpcConnection* conn) > { > - conn->thread = CreateThread(NULL, 0, RPCRT4_io_thread, conn, 0, NULL); > - if (!conn->thread) { > + HANDLE thread = CreateThread(NULL, 0, RPCRT4_io_thread, conn, 0, NULL); > + if (!thread) { > DWORD err = GetLastError(); > ERR("failed to create thread, error=%08lx\n", err); > + RPCRT4_DestroyConnection(conn); > } > + /* we could set conn->thread, but then we'd have to make the io_thread wait > + * for that, otherwise the thread might finish, destroy the connection, and > + * free the memory we'd write to before we did, causing crashes and stuff - > + * so let's implement that later, when we really need conn->thread */ > + > + CloseHandle( thread ); > } > > static DWORD CALLBACK RPCRT4_server_thread(LPVOID the_arg) great, I don't understand this underlying pipe issue yet, but somehow, I'm betting this was just what the doctor ordered to get RPC back in on its feet ... thanks! btw, if you have kludgy patches that you would like to get back to wine, I reccomend you post them to wine-devel, which differentiates "not ready" from "ready" patches (that's just how I do it -- for all I know, this might contradict some official guideline). -- gmt "Democracy is the theory that holds that the common people know what they want, and deserve to get it good and hard." HL Mencken

1 0

Re: remove server silliniess
by John K. Hohm April 18, 2003

April 18, 2003

Thus spake "Gregory M. Turner" <gmturner007(a)ameritech.net>: > ah, the hidden powers of the preprocessor -- how it drives us to obsfucation > with it's robotic stupidity... I wonder, is cpp a turing machine? Makes me > want to implement vim as a sed script ;) I don't think so. However, the C++ compile-time template mechanism *is* a turing machine (proof omitted). That is a persuasive argument, but I'm not sure if it's for or against C++.

1 0

winecfg kickoff
by Dimitrie O. Paun April 17, 2003

April 17, 2003

Hi Alexandre, I have a little (and hopefully not too controversial :) suggesting on how to kickoff our little winecfg utility. Namely, we need to get the load/save from/to registry code going. To do so, we need the regisry key structure in place. Item C1 says "Merge configuration into the registry": http://www.dssd.ca/wine/Wine-0.9-TODO.html#conf So here's my proposal: 1. Do C1 now, but keep the current configuration in place 2. Label the conf file in CVS, to know when we merged it 3. Work on winecfg to load/save stuff properly 4. When done, compare the current version of the conf file with the labeled one, to see what's changed. 5. Adapt winecfg to the new changes 6. Drop the old config file Now, if this sounds like a plan, I was hopeing you could do (1) & (2) as it seems you have an idea where things should end up :) (and because only you can label stuff in CVS). Hopefully people will pick up on (3), if not I guess I can take a look, time permitting. -- Dimi.

2 1

Re: my dmusic and dpnhpast work, revised and resent
by Alexandre Julliard April 17, 2003

April 17, 2003

"Rok Mandeljc" <rokmandeljc(a)hotmail.com> writes: > I just found out that my dpnhpast patch doesn't apply to current cvs tree > (msvcrtd implementation changed Makefile.in and configure). This also mean > that dmusic patch I've sent won't apply either. So I'm sending it again, > this time in one patch. It *should* apply to (today's) cvs cleanly. I have applied all the new dlls, but not the dmusic stubs. I think you should try to group related interfaces in the same source file (and possibly merge some of the existing files too), we can't have 40 different source files for a stub dll. A dozen files is the maximum IMO. -- Alexandre Julliard julliard(a)winehq.com

5 8

Re: Set TZ environment variable when not set
by Rein Klazes April 17, 2003

April 17, 2003

On Wed, 16 Apr 2003 21:05:28 +0200, you wrote: > Changelog: > memory/environ.c: build_environment > Set the TZ value to GMT0, if TZ is not set > > Synplify.exe from the Quicklogic FPGA suite refused to work with the demo > license. Setting TZ to any value made it work. GMT0 should be a good > approach. > > Awaiting the "flame war" about noise level for that silent change :-) Uwe, Some programs, like Agent email/news client that I use, check the existence of TZ before attempting other timezone API call's. With your patch Agent displays all times in UTC, which is not making me happy. A correct patch should take care of the local timezone and daylight saving. Rein. -- Rein Klazes rklazes(a)xs4all.nl

2 4

Re: Initialise tid in WIN_CreateDesktopWindow
by Alexandre Julliard April 17, 2003

April 17, 2003

Uwe Bonnes <bon(a)elektron.ikp.physik.tu-darmstadt.de> writes: > windows/win.c: WIN_CreateDesktopWindow > Initialise pWndDesktop->tid > > This makes the Xilinx command line tools run with ttydrv This is wrong, the desktop should not be owned by the current thread. What's the problem with Xilinx? -- Alexandre Julliard julliard(a)winehq.com

2 1