December 2006 - Wine-Devel - list.winehq.org

Re: oleaut32: Fix failing VarMod tests in WinXP, fix VarMod(VT_DECIMAL) in Wine
by Robert Shearman Dec. 27, 2006

Dec. 27, 2006

a_villacis(a)palosanto.com wrote: > In a preparation for a possible implementation of IDispatch > tests in oleaut32, I decided to clean up the failing tests in > Windows XP: > Great, thanks for fixing this. I was about to do this myself. It's about time we get the number of failing tests on XP as close to zero as possible. -- Rob Shearman

3 2

Re: [Winmm/winealsa] Don't use asynchronous callbacks in dsound any more
by Eric Pouech Dec. 26, 2006

Dec. 26, 2006

Maarten, nice work on Alsa (which is definitively needed !!) a couple of comments: - I know the winmm code (and drivers) is crippled by bad synchronisation tricks (like cleaning a field in a structure to signal a thread). This is bad (TM). So I'd suggest using here a real synchronisation object. - I wonder if it's a good idea to create another thread for this... but before merging all existing drivers' threads into a single one, I'd suggest using the fd oriented functions & poll (see snd_pcm_poll_descriptors and friends) instead of snd_pcm_wait, that'll ease up the future work. Maarten Lankhorst a écrit : > Instead of using asynchronous callbacks that uses signals, use a > seperate thread that can be cancelled, this prevents deadlock issues. > > Basically we use snd_pcm_wait() that tells us when enough room is free > to commit another buffer, then we commit the previous buffer and make > the next buffer ready. > > Since snd_pcm_wait() uses poll(), we don't have signals in winealsa > any more. > ------------------------------------------------------------------------ > > diff --git a/dlls/winmm/winealsa/audio.c b/dlls/winmm/winealsa/audio.c > index 6043680..e1676ed 100644 > --- a/dlls/winmm/winealsa/audio.c > +++ b/dlls/winmm/winealsa/audio.c > @@ -6,6 +6,7 @@ > * Copyright 2002 Eric Pouech > * 2002 Marco Pietrobono > * 2003 Christian Costa : WaveIn support > + * 2006 Maarten Lankhorst > * > * This library is free software; you can redistribute it and/or > * modify it under the terms of the GNU Lesser General Public > @@ -70,11 +71,6 @@ WINE_DEFAULT_DEBUG_CHANNEL(wave); > > #ifdef HAVE_ALSA > > -/* internal ALSALIB functions */ > -/* FIXME: we shouldn't be using internal functions... */ > -snd_pcm_uframes_t _snd_pcm_mmap_hw_ptr(snd_pcm_t *pcm); > - > - > /* state diagram for waveOut writing: > * > * +---------+-------------+---------------+---------------------------------+ > @@ -2627,7 +2623,6 @@ #undef EXIT_ON_ERROR > snd_pcm_hw_params_free(wwo->hw_params); > wwo->hw_params = hw_params; > > - > return wodNotifyClient(wwo, WOM_OPEN, 0L, 0L); > > errexit: > @@ -3045,11 +3040,8 @@ struct IDsDriverBufferImpl > DWORD mmap_buflen_bytes; > snd_pcm_uframes_t mmap_buflen_frames; > snd_pcm_channel_area_t * mmap_areas; > - snd_async_handler_t * mmap_async_handler; > snd_pcm_uframes_t mmap_ppos; /* play position */ > - > - /* Do we have a direct hardware buffer - SND_PCM_TYPE_HW? */ > - int mmap_mode; > + HANDLE mmap_thread; > }; > > static void DSDB_CheckXRUN(IDsDriverBufferImpl* pdbi) > @@ -3076,71 +3068,66 @@ static void DSDB_CheckXRUN(IDsDriverBuff > } > } > > -static void DSDB_MMAPCopy(IDsDriverBufferImpl* pdbi, int mul) > +/** > + * The helper thread for DirectSound > + * > + * Basically it does an infinite loop until it is told to die > + * > + * snd_pcm_wait() is a call that polls the sound buffer and waits > + * until there is at least 1 period free before it returns. > + * > + * We then commit the buffer filled by the owner of this > + * IDSDriverBuffer */ > +static DWORD CALLBACK DBSB_MMAPStart(LPVOID data) > { > + IDsDriverBufferImpl* pdbi = (IDsDriverBufferImpl*)data; > WINE_WAVEDEV * wwo = &(WOutDev[pdbi->drv->wDevID]); > - snd_pcm_uframes_t period_size; > - snd_pcm_sframes_t avail; > - int err; > - int dir=0; > - > + snd_pcm_uframes_t frames, wanted, ofs; > const snd_pcm_channel_area_t *areas; > - snd_pcm_uframes_t ofs; > - snd_pcm_uframes_t frames; > - snd_pcm_uframes_t wanted; > > - if ( !pdbi->mmap_buffer || !wwo->hw_params || !wwo->pcm) > - return; > - > - err = snd_pcm_hw_params_get_period_size(wwo->hw_params, &period_size, &dir); > - avail = snd_pcm_avail_update(wwo->pcm); > + TRACE("\n"); > > - DSDB_CheckXRUN(pdbi); > - > - TRACE("avail=%d, mul=%d\n", (int)avail, mul); > - > - frames = pdbi->mmap_buflen_frames; > - > - EnterCriticalSection(&pdbi->mmap_crst); > + if (areas != pdbi->mmap_areas || areas->addr != pdbi->mmap_areas->addr) > + FIXME("Can't access sound driver's buffer directly.\n"); > > - /* we want to commit the given number of periods, or the whole lot */ > - wanted = mul == 0 ? frames : period_size * 2; > + while (pdbi->mmap_thread) > + { > + snd_pcm_wait(wwo->pcm, -1); > + wanted = frames = pdbi->mmap_buflen_frames; > + DSDB_CheckXRUN(pdbi); > > - snd_pcm_mmap_begin(wwo->pcm, &areas, &ofs, &frames); > - if (areas != pdbi->mmap_areas || areas->addr != pdbi->mmap_areas->addr) > - FIXME("Can't access sound driver's buffer directly.\n"); > - > - /* mark our current play position */ > - pdbi->mmap_ppos = ofs; > - > - if (frames > wanted) > - frames = wanted; > - > - err = snd_pcm_mmap_commit(wwo->pcm, ofs, frames); > - > - /* Check to make sure we committed all we want to commit. ALSA > - * only gives a contiguous linear region, so we need to check this > - * in case we've reached the end of the buffer, in which case we > - * can wrap around back to the beginning. */ > - if (frames < wanted) { > - frames = wanted -= frames; > + EnterCriticalSection(&pdbi->mmap_crst); > snd_pcm_mmap_begin(wwo->pcm, &areas, &ofs, &frames); > snd_pcm_mmap_commit(wwo->pcm, ofs, frames); > - } > > - LeaveCriticalSection(&pdbi->mmap_crst); > + /* mark our current play position */ > + pdbi->mmap_ppos = ofs; > + > + /* Check to make sure we committed all we want to commit. ALSA > + * only gives a contiguous linear region, so we need to check this > + * in case we've reached the end of the buffer, in which case we > + * can wrap around back to the beginning. */ > + if (frames < wanted) { > + frames = wanted - frames; > + snd_pcm_mmap_begin(wwo->pcm, &areas, &ofs, &frames); > + snd_pcm_mmap_commit(wwo->pcm, ofs, frames); > + } > + LeaveCriticalSection(&pdbi->mmap_crst); > + } > + TRACE("Destroyed MMAP thread\n"); > + return 0; > } > > -static void DSDB_PCMCallback(snd_async_handler_t *ahandler) > +/** > + * Cancel running MMAP thread */ > +static void DBSB_MMAPStop(IDsDriverBufferImpl* This) > { > - int periods; > - /* snd_pcm_t * handle = snd_async_handler_get_pcm(ahandler); */ > - IDsDriverBufferImpl* pdbi = snd_async_handler_get_callback_private(ahandler); > - TRACE("callback called\n"); > - > - /* Commit another block (the entire buffer if it's a direct hw buffer) */ > - periods = pdbi->mmap_mode == SND_PCM_TYPE_HW ? 0 : 1; > - DSDB_MMAPCopy(pdbi, periods); > + HANDLE Thread; > + TRACE("Destroying MMAP thread\n"); > + if (This->mmap_thread == NULL) return; > + Thread = This->mmap_thread; > + This->mmap_thread = NULL; > + WaitForSingleObject(Thread, INFINITE); > } > > /** > @@ -3158,21 +3145,47 @@ static int DSDB_CreateMMAP(IDsDriverBuff > unsigned int bits_per_sample; > unsigned int bits_per_frame; > int err; > + int mmap_mode; > + > + mmap_mode = snd_pcm_type(wwo->pcm); > + pdbi->mmap_thread = NULL; > + > + /* If we have a software buffer, change buffer size to only get 2 > + * > + * Why 2? We want a small number so that we don't get ahead of the > + * DirectSound mixer. But we don't want to ever let the buffer get > + * completely empty - having 2 periods gives us time to commit another > + * period when the first expires. > + * > + * The potential for buffer underrun is high, but that's the reality > + * of using a translated buffer (the whole point of DirectSound is > + * to provide direct access to the hardware). > + */ > > - err = snd_pcm_hw_params_get_format(wwo->hw_params, &format); > - err = snd_pcm_hw_params_get_buffer_size(wwo->hw_params, &frames); > - err = snd_pcm_hw_params_get_channels(wwo->hw_params, &channels); > - bits_per_sample = snd_pcm_format_physical_width(format); > - bits_per_frame = bits_per_sample * channels; > - pdbi->mmap_mode = snd_pcm_type(wwo->pcm); > - > - if (pdbi->mmap_mode == SND_PCM_TYPE_HW) { > + if (mmap_mode == SND_PCM_TYPE_HW) { > TRACE("mmap'd buffer is a hardware buffer.\n"); > } > else { > + snd_pcm_uframes_t psize; > + err = snd_pcm_hw_params_get_period_size(wwo->hw_params, &psize, NULL); > + /* Set only a buffer of 2 period sizes, to decrease latency */ > + if (err >= 0) > + err = snd_pcm_hw_params_set_buffer_size_near(wwo->pcm, wwo->hw_params, &psize); > + > + > + psize *= 2; > + if (err < 0) { > + ERR("Errno %d (%s) occured when setting buffer size\n", err, strerror(errno)); > + } > TRACE("mmap'd buffer is an ALSA emulation of hardware buffer.\n"); > } > > + err = snd_pcm_hw_params_get_format(wwo->hw_params, &format); > + err = snd_pcm_hw_params_get_buffer_size(wwo->hw_params, &frames); > + err = snd_pcm_hw_params_get_channels(wwo->hw_params, &channels); > + bits_per_sample = snd_pcm_format_physical_width(format); > + bits_per_frame = bits_per_sample * channels; > + > if (TRACE_ON(wave)) > ALSA_TraceParameters(wwo->hw_params, NULL, FALSE); > > @@ -3208,19 +3221,13 @@ static int DSDB_CreateMMAP(IDsDriverBuff > InitializeCriticalSection(&pdbi->mmap_crst); > pdbi->mmap_crst.DebugInfo->Spare[0] = (DWORD_PTR)"WINEALSA_mmap_crst"; > > - err = snd_async_add_pcm_handler(&pdbi->mmap_async_handler, wwo->pcm, DSDB_PCMCallback, pdbi); > - if ( err < 0 ) > - { > - ERR("add_pcm_handler failed. reason: %s\n", snd_strerror(err)); > - return DSERR_GENERIC; > - } > - > return DS_OK; > } > > static void DSDB_DestroyMMAP(IDsDriverBufferImpl* pdbi) > { > TRACE("mmap buffer %p destroyed\n", pdbi->mmap_buffer); > + DBSB_MMAPStop(pdbi); > pdbi->mmap_areas = NULL; > pdbi->mmap_buffer = NULL; > pdbi->mmap_crst.DebugInfo->Spare[0] = 0; > @@ -3254,6 +3261,7 @@ static ULONG WINAPI IDsDriverBufferImpl_ > > if (refCount) > return refCount; > + IDsDriverBuffer_Stop(iface); > if (This == This->drv->primary) > This->drv->primary = NULL; > DSDB_DestroyMMAP(This); > @@ -3372,29 +3380,13 @@ static HRESULT WINAPI IDsDriverBufferImp > err = snd_pcm_prepare(wwo->pcm); > state = snd_pcm_state(wwo->pcm); > } > - if ( state == SND_PCM_STATE_PREPARED ) > + if ( state == SND_PCM_STATE_PREPARED && This->mmap_thread == NULL) > { > - /* If we have a direct hardware buffer, we can commit the whole lot > - * immediately (periods = 0), otherwise we prime the queue with only > - * 2 periods. > - * > - * Why 2? We want a small number so that we don't get ahead of the > - * DirectSound mixer. But we don't want to ever let the buffer get > - * completely empty - having 2 periods gives us time to commit another > - * period when the first expires. > - * > - * The potential for buffer underrun is high, but that's the reality > - * of using a translated buffer (the whole point of DirectSound is > - * to provide direct access to the hardware). > - * > - * A better implementation would use the buffer Lock() and Unlock() > - * methods to determine how far ahead we can commit, and to rewind if > - * necessary. > - */ > - int periods = This->mmap_mode == SND_PCM_TYPE_HW ? 0 : 2; > - > - DSDB_MMAPCopy(This, periods); > - err = snd_pcm_start(wwo->pcm); > + err = snd_pcm_start(wwo->pcm); > + This->mmap_thread = CreateThread(NULL, 0, DBSB_MMAPStart, (LPVOID)(DWORD)This, 0, NULL); > + if (This->mmap_thread == NULL) > + return DSERR_GENERIC; > + SetThreadPriority(This->mmap_thread, THREAD_PRIORITY_TIME_CRITICAL); > } > return DS_OK; > } > @@ -3419,6 +3411,8 @@ static HRESULT WINAPI IDsDriverBufferImp > return DS_OK; > } > > + DBSB_MMAPStop(This); > + > if ( ( err = snd_pcm_drop(wwo->pcm)) < 0 ) > { > ERR("error while stopping pcm: %s\n", snd_strerror(err)); > > ------------------------------------------------------------------------ > > >

2 2

window managment bug
by Clinton Stimpson Dec. 26, 2006

Dec. 26, 2006

Does bug #6006 fall under window management? When buttons are turned invisible, the invalidated area in the parent window is not repainted, so it looks like the button is still there. Also, there's two invisible buttons that seem to repaint themselves when the parent window resizes, clicking on them does nothing. Any ideas where I can go to track this down? Thanks, Clinton

1 0

Re: oleaut32: preliminary support for VT_DISPATCH as parameters to VarAdd (and other 2-operand operations) (try 3)
by Alexandre Julliard Dec. 26, 2006

Dec. 26, 2006

a_villacis(a)palosanto.com writes: > This is another attempt to add IDispatch support to several 2-operand > functions. Now the macro was removed and the VT_DISPATCH handling was > copied to every affected function. Thanks, it's better that way. Still, it seems to me you don't always clean the variants properly, you should do some sort of "goto end" in all cases instead of simply returning. Also renaming the function parameters in most cases is not necessary. I'd also suggest to do a separate patch for each function, especially since in some functions a bit of restructuring will be needed to cleanup properly on error. -- Alexandre Julliard julliard(a)winehq.org

1 0

dlls/kernel32/tests: test failure in locale.c
by Claudio Fontana Dec. 26, 2006

Dec. 26, 2006

Hello, while tweaking other things in kernel32, I noticed that latest version fails a test in locale.c: $ cd dlls/kernel32/tests $ make test ../../../tools/runtest -q -P wine -M kernel32.dll -T ../../.. -p kernel32_test.exe.so locale.c && touch locale.ok locale.c:1990: Test failed: Enumerated grp locale 1084 not valid locale.c:1990: Test failed: Enumerated grp locale 1170 not valid locale.c:1990: Test failed: Enumerated grp locale 1171 not valid locale.c:1990: Test failed: Enumerated grp locale 2108 not valid locale.c:1990: Test failed: Enumerated grp locale 3132 not valid make: *** [locale.ok] Error 5 Hope this helps. Claudio

2 1

more meaningful traces?
by Lei Zhang Dec. 25, 2006

Dec. 25, 2006

Hi, I noticed several files in dlls/winmm have a getMessage() function to translate message numbers to message names. This is extremely helpful for debug traces, where one can immediately see the name of the messages sent, instead of wondering what 0x105 is. So my question is, why is winmm the only component translating its message? Is there some downside to having message names in the traces? Happy Holidays! - Lei

1 0

Re: [2/2] wined3d: Restore texture bindings in the FBO code
by Stefan Dösinger Dec. 25, 2006

Dec. 25, 2006

Am 25.12.2006 um 17:12 schrieb H. Verbeet: > Previously all textures were (re)bound each DrawPrimitive call. With > the new state management only dirty textures are bound, so we need to > restore the texture binding if we change it. > > Changelog: > - Restore texture bindings in the FBO code > <wined3d_fbo_texture_binding.diff.txt> Why are you not using glPushAttrib / glPopAttrib?

2 1

Re: [comctl32] MonthCalendar - Fix wrong date selected when clicknext/prev months day (#6967)
by Dmitry Timoshkov Dec. 25, 2006

Dec. 25, 2006

"Vijay Kiran Kamuju" <infyquest(a)gmail.com> wrote: > - retval = *daypos + (7 * *weekpos) - firstDay; > + retval = *daypos + ((7 * *weekpos) - firstDay); > return retval; > } This change doesn't make any difference, neither in functionality, nor in readability IMO. Please avoid this kind of changes, it only clutters the patch, making it hard to see what is the real change. > - lpht->st.wDay = MONTHCAL_MonthLength(lpht->st.wMonth,lpht->st.wYear) -day; > + lpht->st.wDay = day+MONTHCAL_MonthLength(lpht->st.wMonth,lpht->st.wYear) ; A couple of spaces around '+' won't hurt, and a space before ';' doesn't match an existing code style. > - return TRUE; > + /*return TRUE;*/ If you are not sure that the return should be removed, it's a sign that you need a test case, just commenting it out without any explanation is not a solution. Same comment to all other similar cases. > + MONTHCAL_CopyTime(&infoPtr->minSel,&nmsc.stSelStart); > + MONTHCAL_CopyTime(&infoPtr->maxSel,&nmsc.stSelEnd); > nmsc.nmhdr.hwndFrom = infoPtr->hwndSelf; > nmsc.nmhdr.idFrom = GetWindowLongPtrW(infoPtr->hwndSelf, GWLP_ID); > nmsc.nmhdr.code = MCN_SELCHANGE; > - MONTHCAL_CopyTime(&infoPtr->minSel,&nmsc.stSelStart); > - MONTHCAL_CopyTime(&infoPtr->maxSel,&nmsc.stSelEnd); This change is not needed. -- Dmitry.

2 2

kernel32 (profile.c): current section and key implementation
by Claudio Fontana Dec. 24, 2006

Dec. 24, 2006

Hello all, I posted previously a patch to wine-patches here: http://www.winehq.org/pipermail/wine-patches/2006-December/034022.html However, I now understand that although it solves the issue I had with the specific program, the change is incomplete, since in several other places, sections and keys are only considered meaningful when not equal to the empty string. In the Microsoft implementation, the unnamed section is valid, as are zero length keys. Thus, I think a broader scope change is required. Before I start working on a new patch, do you acknowledge that the current implementation is incorrect, and a patch fixing it by not requiring name[0] to be != '\0' is likely to be accepted? (btw, I've noticed you switched to git, so I will be using that instead now). Thanks, Claudio Fontana

2 2

Re: ntdll: Added stubs for the Rtl*GenericTable* functions
by Vitaliy Margolen Dec. 24, 2006

Dec. 24, 2006

Is there are any reason why do you need this functions at all? According to ddk they are part of "Installable File System Drivers". This is a kernel level thing only and no user app should use these routines. Especially stabbing them won't do you any good if something actually using them. But that something has no place to run in Wine yet. Vitaliy.

2 5