August 2021 - Wine-Devel - list.winehq.org

[PATCH] dxgi/tests: Add tests to show frame latency object is a semaphore.

by Andrew Eikum

Shin Megami Tensei III Nocturne HD Remaster depends on the frame latency semaphore being released several times after it calls IDXGISwapChain2::SetMaximumFrameLatency(3). Signed-off-by: Andrew Eikum <aeikum(a)codeweavers.com> --- You can also make several Present calls in a row, and the semaphore will be released for each call. However this won't consistently fail in Wine due to timing, so I didn't add those todo_wine tests here. Relevant issue: https://github.com/doitsujin/dxvk/issues/2161 dlls/dxgi/tests/dxgi.c | 18 ++++++++++++++---- 1 file changed, 14 insertions(+), 4 deletions(-) diff --git a/dlls/dxgi/tests/dxgi.c b/dlls/dxgi/tests/dxgi.c index 7a4204c61b5..a83964d7523 100644 --- a/dlls/dxgi/tests/dxgi.c +++ b/dlls/dxgi/tests/dxgi.c @@ -6982,20 +6982,30 @@ static void test_frame_latency_event(IUnknown *device, BOOL is_d3d12) ok(hr == S_OK, "Got unexpected hr %#x.\n", hr); ok(frame_latency == 1, "Got unexpected frame latency %#x.\n", frame_latency); - hr = IDXGISwapChain2_SetMaximumFrameLatency(swapchain2, 2); + hr = IDXGISwapChain2_SetMaximumFrameLatency(swapchain2, 3); ok(hr == S_OK, "Got unexpected hr %#x.\n", hr); hr = IDXGISwapChain2_GetMaximumFrameLatency(swapchain2, &frame_latency); ok(hr == S_OK, "Got unexpected hr %#x.\n", hr); - ok(frame_latency == 2, "Got unexpected frame latency %#x.\n", frame_latency); + ok(frame_latency == 3, "Got unexpected frame latency %#x.\n", frame_latency); + + wait_result = WaitForSingleObject(event, 0); + todo_wine ok(!wait_result, "Got unexpected wait result %#x.\n", wait_result); + wait_result = WaitForSingleObject(event, 0); + todo_wine ok(!wait_result, "Got unexpected wait result %#x.\n", wait_result); + wait_result = WaitForSingleObject(event, 100); + ok(wait_result == WAIT_TIMEOUT, "Got unexpected wait result %#x.\n", wait_result); for (i = 0; i < 5; i++) { hr = IDXGISwapChain2_Present(swapchain2, 0, 0); ok(hr == S_OK, "Present %u failed with hr %#x.\n", i, hr); + + wait_result = WaitForSingleObject(event, 100); + ok(!wait_result, "Got unexpected wait result %#x.\n", wait_result); } - wait_result = WaitForSingleObject(event, 1000); - ok(!wait_result, "Got unexpected wait result %#x.\n", wait_result); + wait_result = WaitForSingleObject(event, 100); + ok(wait_result == WAIT_TIMEOUT, "Got unexpected wait result %#x.\n", wait_result); ref_count = IDXGISwapChain2_Release(swapchain2); ok(!ref_count, "Swap chain has %u references left.\n", ref_count); -- 2.32.0

Aug. 2, 2021

Re: [PATCH v2 1/4] msvcrt/tests: Add test for fopen sequential access hint

by Piotr Caban

Hi Luke, On 8/2/21 5:03 PM, Luke Deller wrote: > +static void test_fopen_hints(void) > +{ > + char *tempf; > + FILE *fp; > + HANDLE handle; > + FILE_MODE_INFORMATION mode_info; > + IO_STATUS_BLOCK io; > + NTSTATUS status; > + > + /* create a test file */ > + tempf = _tempnam(".","wne"); > + fp = fopen(tempf, "wb"); > + ok(fp != NULL, "unable to create test file\n"); > + fwrite("abc\n", 1, 4, fp); > + fclose(fp); > + > + /* open test file with sequential access hint */ > + fp = fopen(tempf, "rbS"); > + ok(fp != NULL, "unable to open test file with sequential access hint\n"); > + > + /* check that sequential access hint is set on underlying file handle */ > + handle = (HANDLE)_get_osfhandle(_fileno(fp)); > + status = NtQueryInformationFile(handle, &io, &mode_info, sizeof(mode_info), > + FileModeInformation); > + ok(!status, "NtQueryInformationFile failed\n"); > + ok(mode_info.Mode & FILE_SEQUENTIAL_ONLY, "handle missing hint\n"); > + > + /* clean up */ > + fclose(fp); > + unlink(tempf); > + free(tempf); > +} It would be nice to also test fp = fopen(tempf, "rbRS"); and fp = fopen(tempf, "rbSR"); case. The same applies to second patch, what happens if _O_RANDOM and _O_SEQUENTIAL is passed? You will also need to mark the tests failing in wine with todo_wine or reorder the patches. Thanks, Piotr

Aug. 2, 2021

[PATCH] wineusb.sys: Remove device from the device list on PDO removal

by Torge Matthies

Fixes a crash when shutting down a prefix. Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=51479 Signed-off-by: Torge Matthies <openglfreak(a)googlemail.com> --- dlls/wineusb.sys/wineusb.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/dlls/wineusb.sys/wineusb.c b/dlls/wineusb.sys/wineusb.c index fae297915fc..28a36b9f380 100644 --- a/dlls/wineusb.sys/wineusb.c +++ b/dlls/wineusb.sys/wineusb.c @@ -413,12 +413,15 @@ static NTSTATUS pdo_pnp(DEVICE_OBJECT *device_obj, IRP *irp) break; case IRP_MN_REMOVE_DEVICE: + EnterCriticalSection(&wineusb_cs); remove_pending_irps(device); libusb_unref_device(device->libusb_device); libusb_close(device->handle); + list_remove(&device->entry); IoDeleteDevice(device->device_obj); + LeaveCriticalSection(&wineusb_cs); ret = STATUS_SUCCESS; break; -- 2.32.0

Aug. 2, 2021

Re: [PATCH v3] dinput/tests: Prove that GetDeviceData() does nothing.

by Rémi Bernon

On 7/27/21 4:49 PM, Arkadiusz Hiler wrote: > Signed-off-by: Arkadiusz Hiler <ahiler(a)codeweavers.com> > --- > > v3: use todo_wine only where applicable > > supersedes 210321 > > dlls/dinput/tests/mouse.c | 47 ++++++++++++++++++++++++++------------- > 1 file changed, 31 insertions(+), 16 deletions(-) > > diff --git a/dlls/dinput/tests/mouse.c b/dlls/dinput/tests/mouse.c > index 69051bb2a34..d4ffcf4853b 100644 > --- a/dlls/dinput/tests/mouse.c > +++ b/dlls/dinput/tests/mouse.c > @@ -106,7 +106,7 @@ static void test_acquire(IDirectInputA *pDI, HWND hwnd) > DIMOUSESTATE m_state; > HWND hwnd2; > DIPROPDWORD di_op; > - DIDEVICEOBJECTDATA mouse_state; > + DIDEVICEOBJECTDATA mouse_state[30]; > DWORD cnt; > int i; > > @@ -159,16 +159,22 @@ static void test_acquire(IDirectInputA *pDI, HWND hwnd) > ok(hr == S_OK, "Acquire() failed: %08x\n", hr); > > mouse_event(MOUSEEVENTF_MOVE, 10, 10, 0, 0); > - cnt = 1; > - hr = IDirectInputDevice_GetDeviceData(pMouse, sizeof(mouse_state), &mouse_state, &cnt, 0); > - ok(hr == S_OK && cnt > 0, "GetDeviceData() failed: %08x cnt:%d\n", hr, cnt); > + cnt = ARRAY_SIZE(mouse_state); > + mouse_state[0].dwTimeStamp = 0xdeadbeef; > + hr = IDirectInputDevice_GetDeviceData(pMouse, sizeof(mouse_state[0]), mouse_state, &cnt, 0); > + ok(hr == S_OK, "GetDeviceData() failed: %08x\n", hr); > + todo_wine ok(cnt == ARRAY_SIZE(mouse_state), "GetDeviceData() unexpected cnt: %d\n", cnt); > + todo_wine ok(mouse_state[0].dwTimeStamp == 0xdeadbeef, "GetDeviceData() overwrote the timestamp: %d\n", mouse_state[0].dwTimeStamp); > > mouse_event(MOUSEEVENTF_MOVE, 10, 10, 0, 0); > hr = IDirectInputDevice_Unacquire(pMouse); > ok(hr == S_OK, "Failed: %08x\n", hr); > - cnt = 1; > - hr = IDirectInputDevice_GetDeviceData(pMouse, sizeof(mouse_state), &mouse_state, &cnt, 0); > - ok(hr == S_OK && cnt > 0, "GetDeviceData() failed: %08x cnt:%d\n", hr, cnt); > + cnt = ARRAY_SIZE(mouse_state); > + mouse_state[0].dwTimeStamp = 0xdeadbeef; > + hr = IDirectInputDevice_GetDeviceData(pMouse, sizeof(mouse_state[0]), mouse_state, &cnt, 0); > + ok(hr == S_OK, "GetDeviceData() failed: %08x\n", hr); > + todo_wine ok(cnt == ARRAY_SIZE(mouse_state), "GetDeviceData() unexpected cnt: %d\n", cnt); > + todo_wine ok(mouse_state[0].dwTimeStamp == 0xdeadbeef, "GetDeviceData() overwrote the timestamp: %d\n", mouse_state[0].dwTimeStamp); > > hr = IDirectInputDevice_Acquire(pMouse); > ok(hr == S_OK, "Failed: %08x\n", hr); > @@ -178,20 +184,29 @@ static void test_acquire(IDirectInputA *pDI, HWND hwnd) > > hr = IDirectInputDevice_Acquire(pMouse); > ok(hr == S_OK, "Failed: %08x\n", hr); > - cnt = 1; > - hr = IDirectInputDevice_GetDeviceData(pMouse, sizeof(mouse_state), &mouse_state, &cnt, 0); > - ok(hr == S_OK && cnt > 0, "GetDeviceData() failed: %08x cnt:%d\n", hr, cnt); > + cnt = ARRAY_SIZE(mouse_state); > + mouse_state[0].dwTimeStamp = 0xdeadbeef; > + hr = IDirectInputDevice_GetDeviceData(pMouse, sizeof(mouse_state[0]), mouse_state, &cnt, 0); > + ok(hr == S_OK, "GetDeviceData() failed: %08x\n", hr); > + todo_wine ok(cnt == ARRAY_SIZE(mouse_state), "GetDeviceData() unexpected cnt: %d\n", cnt); > + todo_wine ok(mouse_state[0].dwTimeStamp == 0xdeadbeef, "GetDeviceData() overwrote the timestamp: %d\n", mouse_state[0].dwTimeStamp); > > /* Check for buffer overflow */ > for (i = 0; i < 6; i++) > mouse_event(MOUSEEVENTF_MOVE, 10 + i, 10 + i, 0, 0); > > - cnt = 1; > - hr = IDirectInputDevice_GetDeviceData(pMouse, sizeof(mouse_state), &mouse_state, &cnt, 0); > - ok(hr == DI_OK, "GetDeviceData() failed: %08x cnt:%d\n", hr, cnt); > - cnt = 1; > - hr = IDirectInputDevice_GetDeviceData(pMouse, sizeof(mouse_state), &mouse_state, &cnt, 0); > - ok(hr == DI_OK && cnt == 1, "GetDeviceData() failed: %08x cnt:%d\n", hr, cnt); > + cnt = ARRAY_SIZE(mouse_state); > + mouse_state[0].dwTimeStamp = 0xdeadbeef; > + hr = IDirectInputDevice_GetDeviceData(pMouse, sizeof(mouse_state[0]), mouse_state, &cnt, 0); > + ok(hr == S_OK, "GetDeviceData() failed: %08x\n", hr); > + todo_wine ok(cnt == ARRAY_SIZE(mouse_state), "GetDeviceData() unexpected cnt: %d\n", cnt); > + todo_wine ok(mouse_state[0].dwTimeStamp == 0xdeadbeef, "GetDeviceData() overwrote the timestamp: %d\n", mouse_state[0].dwTimeStamp); > + cnt = ARRAY_SIZE(mouse_state); > + mouse_state[0].dwTimeStamp = 0xdeadbeef; > + hr = IDirectInputDevice_GetDeviceData(pMouse, sizeof(mouse_state[0]), mouse_state, &cnt, 0); > + ok(hr == S_OK, "GetDeviceData() failed: %08x\n", hr); > + todo_wine ok(cnt == ARRAY_SIZE(mouse_state), "GetDeviceData() unexpected cnt: %d\n", cnt); > + ok(mouse_state[0].dwTimeStamp == 0xdeadbeef, "GetDeviceData() overwrote the timestamp: %d\n", mouse_state[0].dwTimeStamp); > > /* Check for granularity property using BYOFFSET */ > memset(&di_op, 0, sizeof(di_op)); > AFAICS the function does nothing with dinput <= 7, but only when used with DIDEVICEOBJECTDATA struct. It works as expected in the following cases: * dinput >= 8 with size == sizeof(DIDEVICEOBJECTDATA) * dinput <= 7 with size == sizeof(DIDEVICEOBJECTDATA_DX3) The function fails with E_INVALIDARG with dinput >= 8 when called with sizeof(DIDEVICEOBJECTDATA_DX3). I think a few simple tests should validate that (with a test in dinput8/device), but then the tests should instead use the appropriate sizes and check that the event data can be read. Cheers, -- Rémi Bernon <rbernon(a)codeweavers.com>

Aug. 2, 2021

Re: [PATCH v2 2/4] msvcrt: Translate file open access pattern hints

by Marvin

Hi, While running your changed tests, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check? Full results can be found at: https://testbot.winehq.org/JobDetails.pl?Key=94840 Your paranoid android. === debiant2 (32 bit report) === msvcrt: file.c:2760: Test failed: handle missing hint === debiant2 (32 bit Chinese:China report) === msvcrt: file.c:2760: Test failed: handle missing hint === debiant2 (32 bit WoW report) === msvcrt: file.c:2760: Test failed: handle missing hint === debiant2 (64 bit WoW report) === msvcrt: file.c:2760: Test failed: handle missing hint

Aug. 2, 2021

Re: [PATCH v2 1/4] msvcrt/tests: Add test for fopen sequential access hint

by Marvin

Hi, While running your changed tests, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check? Full results can be found at: https://testbot.winehq.org/JobDetails.pl?Key=94839 Your paranoid android. === debiant2 (32 bit report) === msvcrt: file.c:2760: Test failed: handle missing hint === debiant2 (32 bit Arabic:Morocco report) === msvcrt: file.c:2760: Test failed: handle missing hint === debiant2 (32 bit German report) === msvcrt: file.c:2760: Test failed: handle missing hint === debiant2 (32 bit French report) === msvcrt: file.c:2760: Test failed: handle missing hint === debiant2 (32 bit Hebrew:Israel report) === msvcrt: file.c:2760: Test failed: handle missing hint === debiant2 (32 bit Hindi:India report) === msvcrt: file.c:2760: Test failed: handle missing hint === debiant2 (32 bit Japanese:Japan report) === msvcrt: file.c:2760: Test failed: handle missing hint === debiant2 (32 bit Chinese:China report) === msvcrt: file.c:2760: Test failed: handle missing hint === debiant2 (32 bit WoW report) === msvcrt: file.c:2760: Test failed: handle missing hint === debiant2 (64 bit WoW report) === msvcrt: file.c:2760: Test failed: handle missing hint

Aug. 2, 2021

[PATCH v2 4/4] server: Implement file access hints using posix_fadvise

by Luke Deller

Implement file access pattern hints for sequential or random file access using posix_fadvise if available. On Linux this will influence the kernel's readahead behaviour. Signed-off-by: Luke Deller <luke(a)deller.id.au> --- v2: spotted a typo: set the offset argument of posix_fadvise to 0 not 1 (this has no effect on linux, where the call affects the entire file and not just the region specified) --- server/fd.c | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/server/fd.c b/server/fd.c index b953da2ab85..873a94b048a 100644 --- a/server/fd.c +++ b/server/fd.c @@ -2036,6 +2036,14 @@ struct fd *open_fd( struct fd *root, const char *name, struct unicode_str nt_nam free( closed_fd ); fd->cacheable = 1; } + +#if _POSIX_C_SOURCE > 200112L + if (options & FILE_SEQUENTIAL_ONLY) + posix_fadvise( fd->unix_fd, 0, 0, POSIX_FADV_SEQUENTIAL ); + else if (options & FILE_RANDOM_ACCESS) + posix_fadvise( fd->unix_fd, 0, 0, POSIX_FADV_RANDOM ); +#endif + if (root_fd != -1) fchdir( server_dir_fd ); /* go back to the server dir */ return fd; -- 2.25.1

Aug. 2, 2021

[PATCH v2 3/4] kernelbase: Translate FILE_FLAG_SEQUENTIAL_SCAN

by Luke Deller

Translate FILE_FLAG_SEQUENTIAL_SCAN into the corresponding flag for NtCreateFile options Signed-off-by: Luke Deller <luke(a)deller.id.au> --- dlls/kernelbase/file.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/dlls/kernelbase/file.c b/dlls/kernelbase/file.c index e0a75c2ad08..d21ad299d1f 100644 --- a/dlls/kernelbase/file.c +++ b/dlls/kernelbase/file.c @@ -723,6 +723,8 @@ static UINT get_nt_file_options( DWORD attributes ) options |= FILE_SYNCHRONOUS_IO_NONALERT; if (attributes & FILE_FLAG_RANDOM_ACCESS) options |= FILE_RANDOM_ACCESS; + if (attributes & FILE_FLAG_SEQUENTIAL_SCAN) + options |= FILE_SEQUENTIAL_ONLY; if (attributes & FILE_FLAG_WRITE_THROUGH) options |= FILE_WRITE_THROUGH; return options; @@ -3227,6 +3229,7 @@ HANDLE WINAPI DECLSPEC_HOTPATCH OpenFileById( HANDLE handle, LPFILE_ID_DESCRIPTO if (flags & FILE_FLAG_NO_BUFFERING) options |= FILE_NO_INTERMEDIATE_BUFFERING; if (!(flags & FILE_FLAG_OVERLAPPED)) options |= FILE_SYNCHRONOUS_IO_NONALERT; if (flags & FILE_FLAG_RANDOM_ACCESS) options |= FILE_RANDOM_ACCESS; + if (flags & FILE_FLAG_SEQUENTIAL_SCAN) options |= FILE_SEQUENTIAL_ONLY; flags &= FILE_ATTRIBUTE_VALID_FLAGS; objectName.Length = sizeof(ULONGLONG); -- 2.25.1

Aug. 2, 2021

[PATCH v2 2/4] msvcrt: Translate file open access pattern hints

by Luke Deller

Translate access pattern hints supplied to fopen / _open into the corresponding attributes for the lower level CreateFileW call. Signed-off-by: Luke Deller <luke(a)deller.id.au> --- dlls/msvcrt/file.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-) diff --git a/dlls/msvcrt/file.c b/dlls/msvcrt/file.c index 6e955099090..716d397d246 100644 --- a/dlls/msvcrt/file.c +++ b/dlls/msvcrt/file.c @@ -22,8 +22,6 @@ * License along with this library; if not, write to the Free Software * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA * - * TODO - * Use the file flag hints O_SEQUENTIAL, O_RANDOM, O_SHORT_LIVED */ #include <direct.h> @@ -1609,8 +1607,10 @@ static int msvcrt_get_flags(const wchar_t* mode, int *open_flags, int* stream_fl case 'w': break; case 'S': + *open_flags |= _O_SEQUENTIAL; + break; case 'R': - FIXME("ignoring cache optimization flag: %c\n", mode[-1]); + *open_flags |= _O_RANDOM; break; default: ERR("incorrect mode flag: %c\n", mode[-1]); @@ -2269,6 +2269,13 @@ int CDECL _wsopen_dispatch( const wchar_t* path, int oflags, int shflags, int pm sharing |= FILE_SHARE_DELETE; } + if (oflags & _O_RANDOM) + attrib |= FILE_FLAG_RANDOM_ACCESS; + if (oflags & _O_SEQUENTIAL) + attrib |= FILE_FLAG_SEQUENTIAL_SCAN; + if (oflags & _O_SHORT_LIVED) + attrib |= FILE_ATTRIBUTE_TEMPORARY; + sa.nLength = sizeof( SECURITY_ATTRIBUTES ); sa.lpSecurityDescriptor = NULL; sa.bInheritHandle = !(oflags & _O_NOINHERIT); -- 2.25.1

Aug. 2, 2021

[PATCH v2 1/4] msvcrt/tests: Add test for fopen sequential access hint

by Luke Deller

Test that adding "S" to the fopen mode results in the underlying file handle having the sequential access hint set. Signed-off-by: Luke Deller <luke(a)deller.id.au> --- dlls/msvcrt/tests/file.c | 35 +++++++++++++++++++++++++++++++++++ 1 file changed, 35 insertions(+) diff --git a/dlls/msvcrt/tests/file.c b/dlls/msvcrt/tests/file.c index fb242d81282..d02087be46c 100644 --- a/dlls/msvcrt/tests/file.c +++ b/dlls/msvcrt/tests/file.c @@ -35,6 +35,7 @@ #include <process.h> #include <errno.h> #include <locale.h> +#include <winternl.h> #define MSVCRT_FD_BLOCK_SIZE 32 typedef struct { @@ -2731,6 +2732,39 @@ static void test_lseek(void) DeleteFileA("_creat.tst"); } +static void test_fopen_hints(void) +{ + char *tempf; + FILE *fp; + HANDLE handle; + FILE_MODE_INFORMATION mode_info; + IO_STATUS_BLOCK io; + NTSTATUS status; + + /* create a test file */ + tempf = _tempnam(".","wne"); + fp = fopen(tempf, "wb"); + ok(fp != NULL, "unable to create test file\n"); + fwrite("abc\n", 1, 4, fp); + fclose(fp); + + /* open test file with sequential access hint */ + fp = fopen(tempf, "rbS"); + ok(fp != NULL, "unable to open test file with sequential access hint\n"); + + /* check that sequential access hint is set on underlying file handle */ + handle = (HANDLE)_get_osfhandle(_fileno(fp)); + status = NtQueryInformationFile(handle, &io, &mode_info, sizeof(mode_info), + FileModeInformation); + ok(!status, "NtQueryInformationFile failed\n"); + ok(mode_info.Mode & FILE_SEQUENTIAL_ONLY, "handle missing hint\n"); + + /* clean up */ + fclose(fp); + unlink(tempf); + free(tempf); +} + START_TEST(file) { int arg_c; @@ -2803,6 +2837,7 @@ START_TEST(file) test_close(); test__creat(); test_lseek(); + test_fopen_hints(); /* Wait for the (_P_NOWAIT) spawned processes to finish to make sure the report * file contains lines in the correct order -- 2.25.1

Aug. 2, 2021