- Wine-Devel - list.winehq.org

[PATCH v2 8/8] server: Don't allow creating children of orphaned windows.
by Jacek Caban Feb. 8, 2022

Feb. 8, 2022

Signed-off-by: Jacek Caban <jacek(a)codeweavers.com> --- dlls/user32/tests/win.c | 33 +++++++++++++++++++++++++++++++-- server/window.c | 10 +++++++++- 2 files changed, 40 insertions(+), 3 deletions(-)

2 1

[PATCH v2 7/8] server: Orphan child windows when unlinking them from parent.
by Jacek Caban Feb. 8, 2022

Feb. 8, 2022

Signed-off-by: Jacek Caban <jacek(a)codeweavers.com> --- dlls/user32/tests/win.c | 34 ++++++++++++++++++++++++++++++++++ server/window.c | 5 ++++- 2 files changed, 38 insertions(+), 1 deletion(-)

2 1

[PATCH v2 6/8] server: Don't allow orphaned windows in set_parent request.
by Jacek Caban Feb. 8, 2022

Feb. 8, 2022

Signed-off-by: Jacek Caban <jacek(a)codeweavers.com> --- dlls/user32/tests/win.c | 3 --- dlls/user32/win.c | 2 +- server/window.c | 10 +++++++++- 3 files changed, 10 insertions(+), 5 deletions(-)

2 1

[PATCH v2 5/8] user32/tests: Add a test for child window destruction on thread exit.
by Jacek Caban Feb. 8, 2022

Feb. 8, 2022

From: Rémi Bernon <rbernon(a)codeweavers.com> Signed-off-by: Jacek Caban <jacek(a)codeweavers.com> --- dlls/user32/tests/win.c | 235 ++++++++++++++++++++++++++++++++++++++++ 1 file changed, 235 insertions(+)

2 1

[PATCH] ddraw/tests: Skip a crashy test on WARP (v2).
by Stefan Dösinger Feb. 8, 2022

Feb. 8, 2022

ddraw1 doesn't draw, so presumably not affected. Can't find any crashes in the testbot history http://test.winehq.org/data/4364ff8d5c0387edd6fae142affe8e844b291518/win81_… http://test.winehq.org/data/d03984709d87d6eaa0e2e7746f0db5d8fdf81b5a/win150… I haven't found any examples of ddraw7 crashing in this test. Doesn't mean there aren't any. Let's wait and see. Signed-off-by: Stefan Dösinger <stefan(a)codeweavers.com> --- Version 2: Fixed the skip path, somehow releasing the uninitialized surface slipped through. --- dlls/ddraw/tests/ddraw2.c | 20 ++++++++++++++------ dlls/ddraw/tests/ddraw4.c | 11 ++++++++++- 2 files changed, 24 insertions(+), 7 deletions(-) diff --git a/dlls/ddraw/tests/ddraw2.c b/dlls/ddraw/tests/ddraw2.c index 625776c652c..1285fa1b682 100644 --- a/dlls/ddraw/tests/ddraw2.c +++ b/dlls/ddraw/tests/ddraw2.c @@ -1076,6 +1076,17 @@ static void test_coop_level_d3d_state(void) window = create_window(); ddraw = create_ddraw(); ok(!!ddraw, "Failed to create a ddraw object.\n"); + + if (ddraw_is_warp(ddraw)) + { + /* ddraw2 crashes in EndScene, and if it doesn't crash it fails with + * DDERR_SURFACELOST on WARP. */ + win_skip("Skipping test that crashes WARP occasionally.\n"); + IDirectDraw2_Release(ddraw); + DestroyWindow(window); + return; + } + if (!(device = create_device(ddraw, window, DDSCL_NORMAL))) { skip("Failed to create a 3D device, skipping test.\n"); @@ -1159,12 +1170,9 @@ static void test_coop_level_d3d_state(void) hr = IDirect3DDevice2_DrawPrimitive(device, D3DPT_TRIANGLESTRIP, D3DVT_LVERTEX, quad, ARRAY_SIZE(quad), 0); ok(hr == DD_OK, "Got unexpected hr %#x.\n", hr); hr = IDirect3DDevice2_EndScene(device); - ok(hr == DD_OK || broken(ddraw_is_warp(ddraw) && hr == DDERR_SURFACELOST), "Got unexpected hr %#x.\n", hr); - if (hr == DD_OK) - { - color = get_surface_color(rt, 320, 240); - ok(compare_color(color, 0x0000ff80, 1), "Got unexpected color 0x%08x.\n", color); - } + ok(hr == DD_OK, "Got unexpected hr %#x.\n", hr); + color = get_surface_color(rt, 320, 240); + ok(compare_color(color, 0x0000ff80, 1), "Got unexpected color 0x%08x.\n", color); destroy_viewport(device, viewport); destroy_material(background); diff --git a/dlls/ddraw/tests/ddraw4.c b/dlls/ddraw/tests/ddraw4.c index 9b8acb674e3..399ef157497 100644 --- a/dlls/ddraw/tests/ddraw4.c +++ b/dlls/ddraw/tests/ddraw4.c @@ -1301,6 +1301,14 @@ static void test_coop_level_d3d_state(void) hr = IDirect3D3_QueryInterface(d3d, &IID_IDirectDraw4, (void **)&ddraw); ok(hr == DD_OK, "Got unexpected hr %#x.\n", hr); IDirect3D3_Release(d3d); + + if (ddraw_is_warp(ddraw)) + { + /* ddraw4 occasionally crashes in GetRenderTarget. */ + win_skip("Skipping test that crashes WARP occasionally.\n"); + goto done; + } + hr = IDirectDraw4_SetCooperativeLevel(ddraw, window, DDSCL_EXCLUSIVE | DDSCL_FULLSCREEN); ok(hr == DD_OK, "Got unexpected hr %#x.\n", hr); hr = IDirectDrawSurface4_IsLost(rt); @@ -1324,6 +1332,7 @@ static void test_coop_level_d3d_state(void) hr = IDirect3DDevice3_GetRenderTarget(device, &surface); ok(hr == DD_OK, "Got unexpected hr %#x.\n", hr); ok(surface == rt, "Got unexpected surface %p.\n", surface); + IDirectDrawSurface4_Release(surface); hr = IDirect3DDevice3_GetRenderState(device, D3DRENDERSTATE_ZENABLE, &value); ok(hr == DD_OK, "Got unexpected hr %#x.\n", hr); ok(!!value, "Got unexpected z-enable state %#x.\n", value); @@ -1349,8 +1358,8 @@ static void test_coop_level_d3d_state(void) || broken(ddraw_is_warp(ddraw) && compare_color(color, 0x0000ff00, 0)), "Got unexpected color 0x%08x.\n", color); +done: destroy_viewport(device, viewport); - IDirectDrawSurface4_Release(surface); IDirectDrawSurface4_Release(rt); IDirect3DDevice3_Release(device); IDirectDraw4_Release(ddraw); -- 2.34.1

3 2

[PATCH 2/2] winealsa: Remove alsa-pulse handle underrun workaround.
by Huw Davies Feb. 8, 2022

Feb. 8, 2022

Pulse-based systems should be using winepulse.drv. Signed-off-by: Huw Davies <huw(a)codeweavers.com> --- dlls/winealsa.drv/mmdevdrv.c | 102 +---------------------------------- 1 file changed, 1 insertion(+), 101 deletions(-) diff --git a/dlls/winealsa.drv/mmdevdrv.c b/dlls/winealsa.drv/mmdevdrv.c index 8713b6153ce..0c4bd786342 100644 --- a/dlls/winealsa.drv/mmdevdrv.c +++ b/dlls/winealsa.drv/mmdevdrv.c @@ -50,7 +50,6 @@ #include <alsa/asoundlib.h> WINE_DEFAULT_DEBUG_CHANNEL(alsa); -WINE_DECLARE_DEBUG_CHANNEL(winediag); #define NULL_PTR_ERR MAKE_HRESULT(SEVERITY_ERROR, FACILITY_WIN32, RPC_X_NULL_REF_POINTER) @@ -613,89 +612,6 @@ HRESULT WINAPI AUDDRV_GetEndpointIDs(EDataFlow flow, WCHAR ***ids, GUID **guids, return S_OK; } -/* Using the pulse PCM device from alsa-plugins 1.0.24 triggers a bug - * which causes audio to cease playing after a few minutes of playback. - * Setting handle_underrun=1 on pulse-backed ALSA devices seems to work - * around this issue. */ -static snd_config_t *make_handle_underrun_config(const char *name) -{ - snd_config_t *lconf, *dev_node, *hu_node, *type_node; - char dev_node_name[260]; - const char *type_str; - int err; - - snd_config_update(); - - if((err = snd_config_copy(&lconf, snd_config)) < 0){ - WARN("snd_config_copy failed: %d (%s)\n", err, snd_strerror(err)); - return NULL; - } - - sprintf(dev_node_name, "pcm.%s", name); - err = snd_config_search(lconf, dev_node_name, &dev_node); - if(err == -ENOENT){ - snd_config_delete(lconf); - return NULL; - } - if(err < 0){ - snd_config_delete(lconf); - WARN("snd_config_search failed: %d (%s)\n", err, snd_strerror(err)); - return NULL; - } - - /* ALSA is extremely fragile. If it runs into a config setting it doesn't - * recognize, it tends to fail or assert. So we only want to inject - * handle_underrun=1 on devices that we know will recognize it. */ - err = snd_config_search(dev_node, "type", &type_node); - if(err == -ENOENT){ - snd_config_delete(lconf); - return NULL; - } - if(err < 0){ - snd_config_delete(lconf); - WARN("snd_config_search failed: %d (%s)\n", err, snd_strerror(err)); - return NULL; - } - - if((err = snd_config_get_string(type_node, &type_str)) < 0){ - snd_config_delete(lconf); - return NULL; - } - - if(strcmp(type_str, "pulse") != 0){ - snd_config_delete(lconf); - return NULL; - } - - err = snd_config_search(dev_node, "handle_underrun", &hu_node); - if(err >= 0){ - /* user already has an explicit handle_underrun setting, so don't - * use a local config */ - snd_config_delete(lconf); - return NULL; - } - if(err != -ENOENT){ - snd_config_delete(lconf); - WARN("snd_config_search failed: %d (%s)\n", err, snd_strerror(err)); - return NULL; - } - - if((err = snd_config_imake_integer(&hu_node, "handle_underrun", 1)) < 0){ - snd_config_delete(lconf); - WARN("snd_config_imake_integer failed: %d (%s)\n", err, - snd_strerror(err)); - return NULL; - } - - if((err = snd_config_add(dev_node, hu_node)) < 0){ - snd_config_delete(lconf); - WARN("snd_config_add failed: %d (%s)\n", err, snd_strerror(err)); - return NULL; - } - - return lconf; -} - static BOOL get_alsa_name_by_guid(GUID *guid, char *name, DWORD name_size, EDataFlow *flow) { HKEY devices_key; @@ -762,8 +678,6 @@ HRESULT WINAPI AUDDRV_GetAudioEndpoint(GUID *guid, IMMDevice *dev, IAudioClient ACImpl *This; int err; snd_pcm_stream_t stream; - snd_config_t *lconf; - static BOOL handle_underrun = TRUE; char alsa_name[256]; EDataFlow dataflow; HRESULT hr; @@ -800,21 +714,7 @@ HRESULT WINAPI AUDDRV_GetAudioEndpoint(GUID *guid, IMMDevice *dev, IAudioClient } This->dataflow = dataflow; - if(handle_underrun && ((lconf = make_handle_underrun_config(alsa_name)))){ - err = snd_pcm_open_lconf(&This->pcm_handle, alsa_name, stream, SND_PCM_NONBLOCK, lconf); - TRACE("Opening PCM device \"%s\" with handle_underrun: %d\n", alsa_name, err); - snd_config_delete(lconf); - /* Pulse <= 2010 returns EINVAL, it does not know handle_underrun. */ - if(err == -EINVAL){ - ERR_(winediag)("PulseAudio \"%s\" %d without handle_underrun. Audio may hang." - " Please upgrade to alsa_plugins >= 1.0.24\n", alsa_name, err); - handle_underrun = FALSE; - } - }else - err = -EINVAL; - if(err == -EINVAL){ - err = snd_pcm_open(&This->pcm_handle, alsa_name, stream, SND_PCM_NONBLOCK); - } + err = snd_pcm_open(&This->pcm_handle, alsa_name, stream, SND_PCM_NONBLOCK); if(err < 0){ HeapFree(GetProcessHeap(), 0, This); WARN("Unable to open PCM \"%s\": %d (%s)\n", alsa_name, err, snd_strerror(err)); -- 2.25.1

1 0

[PATCH v2 4/8] server: Don't immediately destroy child windows belonging to different thread in free_window_handle.
by Jacek Caban Feb. 8, 2022

Feb. 8, 2022

Notify their thread instead. Signed-off-by: Jacek Caban <jacek(a)codeweavers.com> --- dlls/user32/user_private.h | 18 ------------------ dlls/user32/win.c | 12 +----------- include/ntuser.h | 19 +++++++++++++++++++ server/window.c | 14 +++++++++----- 4 files changed, 29 insertions(+), 34 deletions(-)

1 0

[PATCH v2 3/8] server: Keep a reference to parent in window objects.
by Jacek Caban Feb. 8, 2022

Feb. 8, 2022

Signed-off-by: Jacek Caban <jacek(a)codeweavers.com> --- server/window.c | 29 ++++++++++++++++++++++------- 1 file changed, 22 insertions(+), 7 deletions(-)

1 0

[PATCH v2 2/8] server: Make window struct a server object.
by Jacek Caban Feb. 8, 2022

Feb. 8, 2022

Signed-off-by: Jacek Caban <jacek(a)codeweavers.com> --- Tests show that orphaned windows' parents handles are no longer valid, but a number of operations of children still need to work, even things like MoveWindow(). For that to work, we still need to have parent object around. Following patches will use that for ref-counting, while having full server object is nice for things like leak detection. We could avoid making window object ref-counted by detecting if we're the last child alive when freeing its handle (and then destroying parent), but that doesn't seem cleaner. server/user.h | 2 +- server/window.c | 94 ++++++++++++++++++++++++++++++++++----------- server/winstation.c | 4 +- 3 files changed, 75 insertions(+), 25 deletions(-)

1 0

[PATCH v2 1/8] server: Allocate extra_bytes separatelly from window struct.
by Jacek Caban Feb. 8, 2022

Feb. 8, 2022

Signed-off-by: Jacek Caban <jacek(a)codeweavers.com> --- v2: fixed issues spotted by Huw server/window.c | 22 +++++++++++++++++----- 1 file changed, 17 insertions(+), 5 deletions(-)

1 0