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

[PATCH 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

Re: [PATCH] loader: Add support for ARM linux in the preloader

by Martin Storsjö

On Mon, 2 Aug 2021, Martin Storsjo wrote: > As the preloader is linked with -nodefaultlibs, we need to provide > builtin helpers for division on arm (that normally are provided by > libgcc or compiler-rt). The bundled version is taken from the LLVM > 8.0.0 release, which still was dual licensed under the MIT and > University of Illinois licenses. (In the 9.0.0 release, all of LLVM > was relicensed to a different license.) > > Since 28fe84da45bea7de56539b50eac8ebcec54342de, the main exe image > must be mappable at its desired base address, which essentially > requires the preloader. > > This fixes the original issue that lead to > https://bugs.winehq.org/show_bug.cgi?id=51539. > > Signed-off-by: Martin Storsjo <martin(a)martin.st> > --- I ran into some issues with this patch on some other system, so don't apply it quite yet if it otherwise would have seemed ok, pending some more debugging. // Martin

Aug. 2, 2021

Re: [PATCH vkd3d 7/7] vkd3d: Use Vulkan descriptor arrays if descriptor indexing is available.

by Henri Verbeet

On Fri, 30 Jul 2021 at 06:05, Conor McCarthy <cmccarthy(a)codeweavers.com> wrote: > @@ -2620,7 +2620,8 @@ static bool vk_write_descriptor_set_from_d3d12_desc(VkWriteDescriptorSet *vk_des > case VKD3D_DESCRIPTOR_MAGIC_UAV: > /* We use separate bindings for buffer and texture SRVs/UAVs. > * See d3d12_root_signature_init(). */ > - vk_descriptor_write->dstBinding = vk_binding + 2 * index; > + if(!use_array) > + vk_descriptor_write->dstBinding = vk_binding + 2 * index; Missing space between "if" and "(!use_array)".

Aug. 2, 2021

Re: [PATCH vkd3d 3/7] vkd3d: Return E_INVALIDARG if a descriptor range overflows.

by Henri Verbeet

On Fri, 30 Jul 2021 at 06:04, Conor McCarthy <cmccarthy(a)codeweavers.com> wrote: > @@ -613,6 +613,9 @@ static HRESULT d3d12_root_signature_init_root_descriptor_tables(struct d3d12_roo > if (range->OffsetInDescriptorsFromTableStart != D3D12_DESCRIPTOR_RANGE_OFFSET_APPEND) > offset = range->OffsetInDescriptorsFromTableStart; > > + if (offset + range->NumDescriptors < offset) > + return E_INVALIDARG; > + Zebediah recently introduce the vkd3d_bound_range() helper for this kind of check; we should probably use it here as well.

Aug. 2, 2021

Re: [PATCH vkd3d 2/7] tests: Test for range overflow in test_root_signature_limits().

by Henri Verbeet

On Fri, 30 Jul 2021 at 06:04, Conor McCarthy <cmccarthy(a)codeweavers.com> wrote: > + /* Windows results vary for overflowing to zero, but anything beyond that is invalid. */ > + root_signature_desc.NumParameters = 1; > + descriptor_ranges[0].NumDescriptors = 0x1001; > + descriptor_ranges[0].OffsetInDescriptorsFromTableStart = 0xFFFFF000; Lowercase hexadecimal constants, please.

Aug. 2, 2021

Re: [PATCH vkd3d 6/7] vkd3d: Narrow the scope of cur_binding in d3d12_root_signature_init_root_descriptor_tables().

by Henri Verbeet

Signed-off-by: Henri Verbeet <hverbeet(a)codeweavers.com>

Aug. 2, 2021

Re: [PATCH vkd3d 5/7] vkd3d: Pass target info to vkd3d-shader for compute pipelines.

by Henri Verbeet

Signed-off-by: Henri Verbeet <hverbeet(a)codeweavers.com>

Aug. 2, 2021

Re: [PATCH vkd3d 4/7] vkd3d: Include descriptor indexing in the shader extensions if the device supports it.

by Henri Verbeet

Signed-off-by: Henri Verbeet <hverbeet(a)codeweavers.com>

Aug. 2, 2021

Re: [PATCH vkd3d 1/7] vkd3d: Resolve D3D12_DESCRIPTOR_RANGE_OFFSET_APPEND during table parsing.

by Henri Verbeet

Signed-off-by: Henri Verbeet <hverbeet(a)codeweavers.com>

Aug. 2, 2021

Re: [PATCH v13 3/4] winevulkan: Implement VK_KHR_external_memory_win32 for buffers.

by Derek Lesho

On 7/22/21 1:33 AM, Zebediah Figura (she/her) wrote: > I'm still not convinced that this is the right way to implement this. > What, in specific, is wrong with the way this patch is implemented. In previous mail I mentioned how this patch mostly just sets up the winevulkan infrastructure for converting the pNext chain entries for importing and exporting resources, and accepting it doesn't imply a value-judgement on any potential path forwards in regards to storing metadata or interop with other APIs. Is the problem you have that we use EXTERNAL_MEMORY_fd at all? If I'm not mistaken, we'd be using it in any path we decide to go down. (the wineserver object, whatever it may be, will hold an fd pointing to the underlying resource).

Aug. 2, 2021