March 2017 - Wine-Devel - list.winehq.org

Re: [PATCH 1/6] wined3d: Introduce "copy sub-resource" command stream operation.
by Henri Verbeet March 17, 2017

March 17, 2017

On 17 March 2017 at 12:57, Józef Kucia <jkucia(a)codeweavers.com> wrote: > The idea behind the patch series is to split blit and memcpy-like copy > operations. That wasn't the direction I was going with this though. > A few arguments in favor of this idea: > * memcpy-like is simpler operation (doesn't support conversion or > stretching), Sure, but we do already check for that in wined3d_surface_blt(). > * we shouldn't need to extend blitter to handle other resource types > (it's fine to support just 2D textures), Sure, but extending that interface is trivial, and arguably it should be using texture/sub-resource index anyway. > * blit and copy operations should produce different results between > different formats (blitter performs conversions), That's somewhat of an issue, but we do already have blitter flags for alpha test, color keying, flipping, mirroring, etc. > * copy operation should be allowed between various formats of the same > size (including compressed and non-compressed formats), > * it should be possible to implement wined3d_device_update_texture() > on top of wined3d_cs_emit_copy_sub_resource(). I don't think those are incompatible with the current blitter interface.

1 0

Re: shell32/tests: Return early if the shell folder creation fails in order to avoid a compiler warning.
by Marvin March 17, 2017

March 17, 2017

Hi, While running your changed tests on Windows, 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=29562 Your paranoid android. === wvistau64 (32 bit shlfolder) === shlfolder.c:457: Test failed: 0: got 7480017f expected 7080017f shlfolder.c:457: Test failed: 1: got f480017f expected 7080017f shlfolder.c:457: Test failed: 2: got 44400177 expected 40400177 shlfolder.c:457: Test failed: 3: got 44400177 expected 40400177 shlfolder.c:457: Test failed: 4: got 44400177 expected 40400177 === wvistau64 (64 bit shlfolder) === shlfolder.c:457: Test failed: 0: got 7480017f expected 7080017f shlfolder.c:457: Test failed: 1: got f480017f expected 7080017f shlfolder.c:457: Test failed: 2: got 44400177 expected 40400177 shlfolder.c:457: Test failed: 3: got 44400177 expected 40400177 shlfolder.c:457: Test failed: 4: got 44400177 expected 40400177

2 1

Re: [v4 PATCH] odbccp32: Implement SQLRemoveDriver/SQLRemoveDriverW
by Huw Davies March 17, 2017

March 17, 2017

On Fri, Mar 17, 2017 at 04:22:20AM +0000, Alistair Leslie-Hughes wrote: > v2 - Support Usage Count. > v3 - Write key when not zero. > v4 - Only Close key when it was successfully opened. > > Signed-off-by: Alistair Leslie-Hughes <leslie_alistair(a)hotmail.com> > --- > dlls/odbccp32/odbccp32.c | 69 +++++++++++++++++++++++++++++++++++++++++----- > dlls/odbccp32/tests/misc.c | 4 +-- > 2 files changed, 64 insertions(+), 9 deletions(-) > > diff --git a/dlls/odbccp32/odbccp32.c b/dlls/odbccp32/odbccp32.c > index 037a2a9..5e70a9d 100644 > --- a/dlls/odbccp32/odbccp32.c > +++ b/dlls/odbccp32/odbccp32.c > @@ -1132,22 +1132,77 @@ BOOL WINAPI SQLRemoveDefaultDataSource(void) > return FALSE; > } > > -BOOL WINAPI SQLRemoveDriverW(LPCWSTR lpszDriver, BOOL fRemoveDSN, > - LPDWORD lpdwUsageCount) > +BOOL WINAPI SQLRemoveDriverW(LPCWSTR drivername, BOOL remove_dsn, LPDWORD usage_count) > { > + HKEY hkey; > + DWORD usagecount = 1; > + > clear_errors(); > - FIXME("%s %d %p\n", debugstr_w(lpszDriver), fRemoveDSN, lpdwUsageCount); > - if (lpdwUsageCount) *lpdwUsageCount = 1; > + TRACE("%s %d %p\n", debugstr_w(drivername), remove_dsn, usage_count); > + > + if (RegOpenKeyW(HKEY_LOCAL_MACHINE, odbcini, &hkey) == ERROR_SUCCESS) > + { > + HKEY hkeydriver; > + > + if (RegOpenKeyW(hkey, drivername, &hkeydriver) == ERROR_SUCCESS) > + { > + DWORD size, type; > + > + size = sizeof(usagecount); > + RegGetValueA(hkeydriver, NULL, "UsageCount", RRF_RT_DWORD, &type, &usagecount, &size); > + TRACE("Usage count %d\n", usagecount); > + RegCloseKey(hkeydriver); > + } > + > + if (usagecount) > + usagecount--; > + > + if (!usagecount) > + { > + if (RegDeleteKeyW(hkey, drivername) != ERROR_SUCCESS) > + ERR("Failed to delete registry key: %s\n", debugstr_w(drivername)); > + > + if (RegOpenKeyW(hkey, odbcdrivers, &hkeydriver) == ERROR_SUCCESS) > + { > + if(RegDeleteValueW(hkeydriver, drivername) != ERROR_SUCCESS) > + ERR("Failed to delete registry value: %s\n", debugstr_w(drivername)); > + RegCloseKey(hkeydriver); > + } > + } > + else > + { > + if (RegOpenKeyW(hkey, drivername, &hkeydriver) == ERROR_SUCCESS) > + { > + if (RegSetValueExA(hkeydriver, "UsageCount", 0, REG_DWORD, (BYTE*)&usagecount, sizeof(usagecount)) != ERROR_SUCCESS) > + ERR("Failed to write registry UsageCount key\n"); > + RegCloseKey(hkeydriver); > + } > + } > + > + RegCloseKey(hkey); > + } > + It should be possible to do this without needing to open the drivername key twice. Huw.

1 0

Re: [PATCH v2] user32: Always send WM_NCACTIVATE when changing the foreground window.
by Alexandre Julliard March 17, 2017

March 17, 2017

Alex Henrie <alexhenrie24(a)gmail.com> writes: > Fixes https://bugs.winehq.org/show_bug.cgi?id=7198 > > v2: Use the wineserver to determine if the new foreground window is > already active in its thread. It still looks suspicious to do it here, particularly since we already have code for this in set_active_window. It would need more tests to determine where it fits with the other messages of the sequence. -- Alexandre Julliard julliard(a)winehq.org

2 3

Re: ddraw/tests: Fix Android build by replacing log2f(x) with logf(x)/logf(2)
by Alexandre Julliard March 16, 2017

March 16, 2017

André Hentschel <nerv(a)dawncrow.de> writes: > Signed-off-by: André Hentschel <nerv(a)dawncrow.de> > --- > dlls/ddraw/tests/ddraw1.c | 2 +- > dlls/ddraw/tests/ddraw2.c | 2 +- > dlls/ddraw/tests/ddraw4.c | 2 +- > dlls/ddraw/tests/ddraw7.c | 2 +- > 4 files changed, 4 insertions(+), 4 deletions(-) It builds fine here, are you using an old API version? -- Alexandre Julliard julliard(a)winehq.org

2 1

Re: [v2 2/6] d3dx9: Apply changed states only in CommitChanges.
by Matteo Bruni March 16, 2017

March 16, 2017

2017-03-14 13:46 GMT+01:00 Paul Gofman <gofmanp(a)gmail.com>: > Signed-off-by: Paul Gofman <gofmanp(a)gmail.com> The current subject might be slightly confusing, something like "d3dx9: Don't apply unmodified states in CommitChanges." would be probably better. The patch looks generally good. Comments follow. > --- > v2: > - stored current_light and current_material in effect structure instead of pass; > - renamed clear_dirty_all() -> clear_dirty_params(). > --- > +static BOOL param_clear_dirty_func(void *dummy, struct d3dx_parameter *param) > +{ > + *param->dirty_flag_ptr &= ~PARAMETER_FLAG_DIRTY; > + return FALSE; > +} > + > +static void clear_dirty_params(struct d3dx9_base_effect *base) > +{ > + unsigned int i; > + > + for (i = 0; i < base->parameter_count; ++i) > + walk_parameter_tree(&base->parameters[i], param_clear_dirty_func, NULL); > +} The walk seems unnecessary since we know that all the children parameters share their dirty flag with their parents. > - param = param->referenced_param; > - TRACE("Array index %u.\n", array_idx); > - if (array_idx >= param->element_count) > + if (array_idx >= ref_param->element_count) > { > - ERR("Computed array index %u is out of bound %u.\n", array_idx, param->element_count); > + ERR("Computed array index %u is out of bound %u.\n", array_idx, ref_param->element_count); > return D3DERR_INVALIDCALL; > } > - param = &param->members[array_idx]; > + param = &ref_param->members[array_idx]; > + *param_dirty = state->index != array_idx || is_param_dirty(param); This looked a bit weird to me at first because of reusing "param". Maybe use a separate "selected_param" variable (or something to that effect)? > - if (FAILED(hr = d3dx9_get_param_value_ptr(effect, pass, state, &param_value, &param))) > + if (FAILED(hr = d3dx9_get_param_value_ptr(effect, pass, state, &param_value, &param, > + update_all, &param_dirty))) > /* Native d3dx returns D3D_OK from BeginPass or Commit involving out of bounds array > - * access and does not touch affected state. */ > + * access and does not touch affected state, except for BeginPass fails with E_FAIL if > + * out of bounds array index depends on dirty parameter(s) in BeginPass. The latter case > + * is TODO. */ Nitpicking on the comment, it sounds better to me with something like "...except for BeginPass when the out of bounds array index depends on dirty parameters. The latter case is supposed to return E_FAIL but is currently TODO". BTW, the ERR in d3dx9_get_param_value_ptr() should probably be a WARN since it can be triggered by the application. > + TRACE("d3dx9_get_param_value_ptr error %#x.\n", hr); This one shouldn't be necessary, I think there should have been a more specific error / warning printed out by d3dx9_get_param_value_ptr() or one of its callees. Otherwise you could repurpose this to say that you're returning D3D_OK notwithstanding the failure. > + return D3D_OK; > + } Actually, it's probably better to change the d3dx9_get_param_value_ptr() return on out of bounds to E_FAIL and only override the return value to D3D_OK in that specific case. > +BOOL is_param_eval_input_dirty(struct d3dx_param_eval *peval) > +{ > + return is_const_tab_input_dirty(&peval->pres.inputs) > + || is_const_tab_input_dirty(&peval->shader_inputs); > +} If I'm not mistaken this only needs to check pres.inputs, given that it's only reasonably going to be called on FXLC and ARRAY_SELECTOR preshaders (not on actual shaders' preshaders). It doesn't matter much though. > - set_constants(rs, &pres->inputs); > - if (FAILED(hr = execute_preshader(pres))) > - return hr; > + if (update_all || is_const_tab_input_dirty(&pres->inputs)) > + { > + set_constants(rs, &pres->inputs, TRUE); > + if (FAILED(hr = execute_preshader(pres))) > + return hr; > + regstore_reset_table(rs, PRES_REGTAB_CONST); > + } This is technically a separate bugfix which would be better as a separate patch, before this one.

2 6

Re: [PATCH 2/2] rpcrt4/tests: Added more server listening tests.
by Marvin March 16, 2017

March 16, 2017

Hi, While running your changed tests on Windows, 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=29547 Your paranoid android. === w8 (32 bit server) === server.c:1769: Test failed: RpcServerUseProtseqEp(ncacn_np) failed with status 1740 server.c:1776: Test failed: RpcServerListen failed with status 1714 server.c:1777: Test failed: RpcMgmtIsServerListening returned 1715, expected 0 server.c:1780: Test failed: RpcServerListen failed with status 1714 server.c:1783: Test failed: RpcMgmtStopServerListening server.c:1787: Test failed: RpcMgmtWaitServerListening failed with status 1715 === w864 (32 bit server) === server.c:1769: Test failed: RpcServerUseProtseqEp(ncacn_np) failed with status 1740 server.c:1776: Test failed: RpcServerListen failed with status 1714 server.c:1777: Test failed: RpcMgmtIsServerListening returned 1715, expected 0 server.c:1780: Test failed: RpcServerListen failed with status 1714 server.c:1783: Test failed: RpcMgmtStopServerListening server.c:1787: Test failed: RpcMgmtWaitServerListening failed with status 1715 === w1064 (32 bit server) === server.c:1769: Test failed: RpcServerUseProtseqEp(ncacn_np) failed with status 1740 server.c:1776: Test failed: RpcServerListen failed with status 1714 server.c:1777: Test failed: RpcMgmtIsServerListening returned 1715, expected 0 server.c:1780: Test failed: RpcServerListen failed with status 1714 server.c:1783: Test failed: RpcMgmtStopServerListening server.c:1787: Test failed: RpcMgmtWaitServerListening failed with status 1715 === w864 (64 bit server) === server.c:1769: Test failed: RpcServerUseProtseqEp(ncacn_np) failed with status 1740 server.c:1776: Test failed: RpcServerListen failed with status 1714 server.c:1777: Test failed: RpcMgmtIsServerListening returned 1715, expected 0 server.c:1780: Test failed: RpcServerListen failed with status 1714 server.c:1783: Test failed: RpcMgmtStopServerListening server.c:1787: Test failed: RpcMgmtWaitServerListening failed with status 1715 === w1064 (64 bit server) === server.c:1769: Test failed: RpcServerUseProtseqEp(ncacn_np) failed with status 1740 server.c:1776: Test failed: RpcServerListen failed with status 1714 server.c:1777: Test failed: RpcMgmtIsServerListening returned 1715, expected 0 server.c:1780: Test failed: RpcServerListen failed with status 1714 server.c:1783: Test failed: RpcMgmtStopServerListening server.c:1787: Test failed: RpcMgmtWaitServerListening failed with status 1715

1 0

Re: [PATCH] wined3d: Zero-initialize count.
by Henri Verbeet March 16, 2017

March 16, 2017

On 15 March 2017 at 17:50, Huw Davies <huw(a)codeweavers.com> wrote: > This prevents the loop below accessing beyond multisample_types > if the GL_NUM_SAMPLE_COUNTS call fails. > I think making this a little more robust is fine, but that call should never fail.

1 0

Re: [PATCH 4/8] wined3d: Allow using more than MAX_COMBINED_SAMPLERS texture image units.
by Matteo Bruni March 16, 2017

March 16, 2017

2017-03-14 13:15 GMT+01:00 Józef Kucia <jkucia(a)codeweavers.com>: > Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> > diff --git a/dlls/wined3d/utils.c b/dlls/wined3d/utils.c > index b1cda46..6fe96e4 100644 > --- a/dlls/wined3d/utils.c > +++ b/dlls/wined3d/utils.c > @@ -5994,40 +5994,28 @@ void wined3d_gl_limits_get_uniform_block_range(const struct wined3d_gl_limits *g > void wined3d_gl_limits_get_texture_unit_range(const struct wined3d_gl_limits *gl_limits, > enum wined3d_shader_type shader_type, unsigned int *base, unsigned int *count) > { > - static const struct > + if (shader_type == WINED3D_SHADER_TYPE_COMPUTE) > { > - enum wined3d_shader_type type; > - unsigned int base_idx; > - unsigned int count; > + if (gl_limits->combined_samplers == gl_limits->graphics_samplers) > + *base = 0; > + else > + *base = gl_limits->graphics_samplers - 1; > + *count = gl_limits->compute_samplers; > + return; > } Late to the party and not really important but shouldn't *base in the else case be set to gl_limits->graphics_samplers (i.e. without the "- 1")?

2 1

Re: [v3 PATCH] odbccp32: Implement SQLRemoveDriver/SQLRemoveDriverW
by Huw Davies March 16, 2017

March 16, 2017

On Thu, Mar 16, 2017 at 04:54:08AM +0000, Alistair Leslie-Hughes wrote: > v2 - Support Usage Count. > v3 - Write key when not zero. > > Signed-off-by: Alistair Leslie-Hughes <leslie_alistair(a)hotmail.com> > --- > dlls/odbccp32/odbccp32.c | 66 +++++++++++++++++++++++++++++++++++++++++----- > dlls/odbccp32/tests/misc.c | 4 +-- > 2 files changed, 61 insertions(+), 9 deletions(-) > > diff --git a/dlls/odbccp32/odbccp32.c b/dlls/odbccp32/odbccp32.c > index 037a2a9..ace801f 100644 > --- a/dlls/odbccp32/odbccp32.c > +++ b/dlls/odbccp32/odbccp32.c > @@ -1132,22 +1132,74 @@ BOOL WINAPI SQLRemoveDefaultDataSource(void) > return FALSE; > } > > -BOOL WINAPI SQLRemoveDriverW(LPCWSTR lpszDriver, BOOL fRemoveDSN, > - LPDWORD lpdwUsageCount) > +BOOL WINAPI SQLRemoveDriverW(LPCWSTR drivername, BOOL remove_dsn, LPDWORD usage_count) > { > + HKEY hkey; > + DWORD usagecount = 1; > + > clear_errors(); > - FIXME("%s %d %p\n", debugstr_w(lpszDriver), fRemoveDSN, lpdwUsageCount); > - if (lpdwUsageCount) *lpdwUsageCount = 1; > + TRACE("%s %d %p\n", debugstr_w(drivername), remove_dsn, usage_count); > + > + if (RegOpenKeyW(HKEY_LOCAL_MACHINE, odbcini, &hkey) == ERROR_SUCCESS) > + { > + HKEY hkeydriver; > + > + if (RegOpenKeyW(hkey, drivername, &hkeydriver) == ERROR_SUCCESS) > + { > + DWORD size, type; > + > + size = sizeof(usagecount); > + RegGetValueA(hkeydriver, NULL, "UsageCount", RRF_RT_DWORD, &type, &usagecount, &size); > + TRACE("Usage count %d\n", usagecount); > + } > + > + if (usagecount) > + usagecount--; > + > + if (!usagecount) > + { > + RegCloseKey(hkeydriver); > + > + if (RegDeleteKeyW(hkey, drivername) != ERROR_SUCCESS) > + ERR("Failed to delete registry key: %s\n", debugstr_w(drivername)); > + > + if (RegOpenKeyW(hkey, odbcdrivers, &hkeydriver) == ERROR_SUCCESS) > + { > + if(RegDeleteValueW(hkeydriver, drivername) != ERROR_SUCCESS) > + ERR("Failed to delete registry value: %s\n", debugstr_w(drivername)); > + } > + } > + else > + { > + if (RegSetValueExA(hkeydriver, "UsageCount", 0, REG_DWORD, (BYTE*)&usagecount, sizeof(usagecount)) != ERROR_SUCCESS) > + ERR("Failed to write registry UsageCount key\n"); > + } > + RegCloseKey(hkeydriver); > + > + RegCloseKey(hkey); > + } > + > + if (usage_count) > + *usage_count = usagecount; > + > return TRUE; > } This isn't correct if RegOpenKeyW(hkey, drivername, &hkeydriver) fails. The clue that something is wrong (at least for this style of handling errors) is the two RegCloseKey()s in the same block. Huw.

1 0