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

Re: [PATCH] Made ODBC function signatures 64-bit compliant
by Alistair Leslie-Hughes Feb. 15, 2017

Feb. 15, 2017

On 15/02/17 05:55, Andreas Damm wrote: > From: Andreas Damm <andreas(a)insightsoftware.com> > > --- > dlls/odbc32/proxyodbc.c | 108 +++++++++++++++++++++++----------------------- > include/sql.h | 36 ++++++++-------- > include/sqlext.h | 22 +++++----- > 3 files changed, 83 insertions(+), 83 deletions(-) ~/wine$ patch -p1 < \[PATCH\]\ Made\ ODBC\ function\ signatures\ 64-bit\ compliant.txt This fails to apply against git. patching file dlls/odbc32/proxyodbc.c Hunk #11 succeeded at 1095 with fuzz 2. Hunk #14 succeeded at 1408 with fuzz 2. Hunk #15 succeeded at 1420 with fuzz 2. Hunk #16 succeeded at 1445 with fuzz 2. Hunk #19 succeeded at 1546 with fuzz 2. Hunk #29 succeeded at 2412 with fuzz 2. patching file include/sql.h Hunk #1 FAILED at 506. Hunk #7 FAILED at 669. 2 out of 11 hunks FAILED -- saving rejects to file include/sql.h.rej patching file include/sqlext.h Hunk #1 FAILED at 1455. Hunk #2 FAILED at 1472. Hunk #3 FAILED at 1515. Hunk #5 FAILED at 1578. Hunk #6 succeeded at 1629 with fuzz 1. 4 out of 6 hunks FAILED -- saving rejects to file include/sqlext.h.rej Best Regards Alistair Leslie-Hughes

3 2

Re: [3/4] Use CP 1200 (Unicode) for Win32/Win64 version info.
by Alexandre Julliard Feb. 14, 2017

Feb. 14, 2017

Akihiro Sagawa <sagawa.aki(a)gmail.com> writes: > Signed-off-by: Akihiro Sagawa <sagawa.aki(a)gmail.com> > --- > dlls/version/tests/info.c | 2 -- > include/wine/wine_common_ver.rc | 10 +++++++++- > 2 files changed, 9 insertions(+), 3 deletions(-) This is breaking some tests: ../../../tools/runtest -q -P wine -T ../../.. -M mshtml.dll -p mshtml_test.exe.so activex && touch activex.ok activex.c:2426: Test failed: expected CreateInstance activex.c:2429: Test failed: expected QuickActivate activex.c:2432: Test failed: expected IPersistPropertyBag_Load activex.c:2433: Test failed: expected Invoke_READYSTATE activex.c:2434: Test failed: expected Invoke_SECURITYCTX activex.c:2435: Test failed: expected Invoke_SCRIPTCALL activex.c:2440: Test failed: expected DoVerb activex.c:2443: Test failed: expected InPlaceObject_GetWindow activex.c:2444: Test failed: expected SetObjectRects wine: Unhandled page fault on read access to 0x00000000 at address 0x7ecb5d28 (thread 0072), starting debugger... -- Alexandre Julliard julliard(a)winehq.org

1 0

Re: [PATCH] Made ODBC function signatures 64-bit compliant
by Alistair Leslie-Hughes Feb. 14, 2017

Feb. 14, 2017

On 15/02/17 05:55, Andreas Damm wrote: > From: Andreas Damm <andreas(a)insightsoftware.com> > > --- > dlls/odbc32/proxyodbc.c | 108 +++++++++++++++++++++++----------------------- > include/sql.h | 36 ++++++++-------- > include/sqlext.h | 22 +++++----- > 3 files changed, 83 insertions(+), 83 deletions(-) ~/wine$ patch -p1 < \[PATCH\]\ Made\ ODBC\ function\ signatures\ 64-bit\ compliant.txt This fails to apply against git. patching file dlls/odbc32/proxyodbc.c Hunk #11 succeeded at 1095 with fuzz 2. Hunk #14 succeeded at 1408 with fuzz 2. Hunk #15 succeeded at 1420 with fuzz 2. Hunk #16 succeeded at 1445 with fuzz 2. Hunk #19 succeeded at 1546 with fuzz 2. Hunk #29 succeeded at 2412 with fuzz 2. patching file include/sql.h Hunk #1 FAILED at 506. Hunk #7 FAILED at 669. 2 out of 11 hunks FAILED -- saving rejects to file include/sql.h.rej patching file include/sqlext.h Hunk #1 FAILED at 1455. Hunk #2 FAILED at 1472. Hunk #3 FAILED at 1515. Hunk #5 FAILED at 1578. Hunk #6 succeeded at 1629 with fuzz 1. 4 out of 6 hunks FAILED -- saving rejects to file include/sqlext.h.rej Best Regards Alistair Leslie-Hughes

1 0

Re: [v3 PATCH] d2d1: Added support for ID2D1GdiInteropRenderTarget interface
by Jacek Caban Feb. 14, 2017

Feb. 14, 2017

Hi Nikolay, It looks correct to me now. On 14.02.2017 16:19, Nikolay Sivov wrote: > --- a/dlls/d2d1/bitmap_render_target.c > +++ b/dlls/d2d1/bitmap_render_target.c > @@ -32,6 +32,8 @@ static inline struct d2d_bitmap_render_target *impl_from_ID2D1BitmapRenderTarget > static HRESULT STDMETHODCALLTYPE d2d_bitmap_render_target_QueryInterface(ID2D1BitmapRenderTarget *iface, > REFIID iid, void **out) > { > + struct d2d_bitmap_render_target *render_target = impl_from_ID2D1BitmapRenderTarget(iface); > + > TRACE("iface %p, iid %s, out %p.\n", iface, debugstr_guid(iid), out); > > if (IsEqualGUID(iid, &IID_ID2D1BitmapRenderTarget) > @@ -43,6 +45,8 @@ static HRESULT STDMETHODCALLTYPE d2d_bitmap_render_target_QueryInterface(ID2D1Bi > *out = iface; > return S_OK; > } > + else if (IsEqualGUID(iid, &IID_ID2D1GdiInteropRenderTarget)) > + return ID2D1RenderTarget_QueryInterface(render_target->dxgi_target, iid, out); You could just call QI on dxgi_target unconditionally here. It will handle failure for you as well. Also by having this logic in one place, if there are any more interfaces that you'd like to expose from dxgi_target in the future, only dxgi_target QI change would be needed. That's not very important at the moment and I don't really know relevant code, so feel free to ignore that comment :) Thanks, Jacek

2 1

Re: [4/4] version: Replace kernel32 translation info in non-US locales.
by Alexandre Julliard Feb. 14, 2017

Feb. 14, 2017

Akihiro Sagawa <sagawa.aki(a)gmail.com> writes: > @@ -1078,6 +1086,67 @@ BOOL WINAPI VerQueryValueW( LPCVOID pBlock, LPCWSTR lpSubBlock, > return VersionInfo32_QueryValue(info, lpSubBlock, lplpBuffer, puLen, NULL); > } > > +/* replace translation info as some apps requires localised info from kernel32 */ > +static void replace_translation_info( void *info, DWORD len ) > +{ > + static const WCHAR product_nameW[] = {'\\','S','t','r','i','n','g', > + 'F','i','l','e','I','n','f','o', > + '\\','0','4','0','9','0','4','B','0', > + '\\','P','r','o','d','u','c','t', > + 'N','a','m','e',0}; > + static const WCHAR original_filenameW[] = {'\\','S','t','r','i','n','g', > + 'F','i','l','e','I','n','f','o', > + '\\','0','4','0','9','0','4','B','0', > + '\\','O','r','i','g','i','n','a','l', > + 'F','i','l','e','n','a','m','e',0}; > + static const WCHAR wineW[] = {'W','i','n','e', 0}; > + static const WCHAR kernel32W[] = {'k','e','r','n','e','l','3','2','.','d','l','l',0}; That's very hackish, we need a more general solution. -- Alexandre Julliard julliard(a)winehq.org

2 3

Re: [PATCH 3/9] server: Allow creating async objects without associated queue.
by Alexandre Julliard Feb. 14, 2017

Feb. 14, 2017

Jacek Caban <jacek(a)codeweavers.com> writes: > Creating async object also validates and resets passed event, which (at > least in some cases) should be done too early to queue the async. That's > what following patches do for read and write to match error handling > pattern. > > Also, it seems nice to pass ready to use async object instead of > async_data_t pointer to fd ops. Maybe, but then please clean this up some more. Instead of adding a bunch of optional arguments, please split the functionality into separate functions. For instance, if create_async() is not always queuing, then make queuing a separate step in all cases. Also if we pass the async object we shouldn't need to pass the async data or iosb. -- Alexandre Julliard julliard(a)winehq.org

1 0

Re: [PATCH v5 1/3] storage.dll16: Simplify operations in IStream16::Seek.
by Alexandre Julliard Feb. 14, 2017

Feb. 14, 2017

Zebediah Figura <z.figura12(a)gmail.com> writes: > + case SEEK_END: > + if (offset.QuadPart > This->stde.pps_size) > + return STG_E_INVALIDFUNCTION; > + > + This->offset.QuadPart = This->stde.pps_size-offset.QuadPart; I know the previous code was doing this too, but it doesn't look right. The offset should be added, not subtracted. > + /* offset must be ==0 (<0 is invalid, and >0 cannot be handled > + * right now. > + */ > + assert(This->offset.u.HighPart == 0); You should get rid of the assert while you are at it, that's not appropriate for parameters passed from the application. -- Alexandre Julliard julliard(a)winehq.org

1 0

Re: [PATCH 1/5] crypt32: Test CryptStringToBinary with weird base64.
by Alexandre Julliard Feb. 14, 2017

Feb. 14, 2017

Lauri Kenttä <lauri.kentta(a)gmail.com> writes: > +static void decodeBase64WithLenFmt(LPCSTR str, int len, DWORD fmt, LPCSTR expected, int le, BOOL isBroken) > +{ > + BYTE buf[8] = {0}; > + DWORD bufLen = sizeof(buf), fmtUsed; > + BOOL ret; > + SetLastError(0); Something like 0xdeadbeef would be better. > + ret = pCryptStringToBinaryA(str, len, fmt, buf, &bufLen, NULL, &fmtUsed); > + buf[bufLen] = 0; This will overflow the buffer if the length hasn't been set. -- Alexandre Julliard julliard(a)winehq.org

1 0

Re: [PATCH 3/4 v2] gdiplus: Reimplement GdipCreateHBITMAPFromBitmap so it can work on locked bitmaps
by Vincent Povirk Feb. 14, 2017

Feb. 14, 2017

+ GdipGetImageFlags(&bitmap->image, &flags); The result here is unused.

1 0

Re: [v4 1/4] ntdll: Proper handling of debug regs on x86-64.
by Sebastian Lackner Feb. 13, 2017

Feb. 13, 2017

On 12.02.2017 23:41, Andrew Wesie wrote: > Based on existing code for i386. Code is shared when possible to avoid too > much duplication, even though this leads to more ifdefs. > > Signed-off-by: Andrew Wesie <awesie(a)gmail.com> > --- > dlls/ntdll/ntdll_misc.h | 28 ++++++++++---------- > dlls/ntdll/signal_x86_64.c | 64 +++++++++++++++++++++++++++++++++++++++++++--- > dlls/ntdll/thread.c | 12 +++++++-- > 3 files changed, 85 insertions(+), 19 deletions(-) > > diff --git a/dlls/ntdll/ntdll_misc.h b/dlls/ntdll/ntdll_misc.h > index 5e4c39e..5c5cbd2 100644 > --- a/dlls/ntdll/ntdll_misc.h > +++ b/dlls/ntdll/ntdll_misc.h > @@ -215,25 +215,27 @@ struct debug_info > /* thread private data, stored in NtCurrentTeb()->SpareBytes1 */ > struct ntdll_thread_data > { > +#if defined(__i386__) || defined(__x86_64__) > + DWORD_PTR dr0; /* 1bc/2e8 Debug registers */ > + DWORD_PTR dr1; /* 1c0/2f0 */ > + DWORD_PTR dr2; /* 1c4/2f8 */ > + DWORD_PTR dr3; /* 1c8/300 */ > + DWORD_PTR dr6; /* 1cc/308 */ > + DWORD_PTR dr7; /* 1d0/310 */ > +#endif > #ifdef __i386__ > - DWORD dr0; /* 1bc Debug registers */ > - DWORD dr1; /* 1c0 */ > - DWORD dr2; /* 1c4 */ > - DWORD dr3; /* 1c8 */ > - DWORD dr6; /* 1cc */ > - DWORD dr7; /* 1d0 */ > DWORD fs; /* 1d4 TEB selector */ > DWORD gs; /* 1d8 libc selector; update winebuild if you move this! */ > void *vm86_ptr; /* 1dc data for vm86 mode */ > #else > - void *exit_frame; /* /2e8 exit frame pointer */ > + void *exit_frame; /* /318 exit frame pointer */ > #endif > - struct debug_info *debug_info; /* 1e0/2f0 info for debugstr functions */ > - int request_fd; /* 1e4/2f8 fd for sending server requests */ > - int reply_fd; /* 1e8/2fc fd for receiving server replies */ > - int wait_fd[2]; /* 1ec/300 fd for sleeping server requests */ > - BOOL wow64_redir; /* 1f4/308 Wow64 filesystem redirection flag */ > - pthread_t pthread_id; /* 1f8/310 pthread thread id */ > + struct debug_info *debug_info; /* 1e0/320 info for debugstr functions */ > + int request_fd; /* 1e4/328 fd for sending server requests */ > + int reply_fd; /* 1e8/32c fd for receiving server replies */ > + int wait_fd[2]; /* 1ec/330 fd for sleeping server requests */ > + BOOL wow64_redir; /* 1f4/338 Wow64 filesystem redirection flag */ > + pthread_t pthread_id; /* 1f8/340 pthread thread id */ > #ifdef __i386__ > WINE_VM86_TEB_INFO vm86; /* 1fc vm86 private data */ > void *exit_frame; /* 204 exit frame pointer */ Adding those fields should work, but it is a bit dangerous because we only have limited space. I would suggest adding asserts to ensure we never make this struct too big. Something like this should work: C_ASSERT( FIELD_OFFSET(TEB, SpareBytes1) + sizeof(struct ntdll_thread_data) <= FIELD_OFFSET(TEB, GdiTebBatch) + sizeof(((TEB *)0)->GdiTebBatch) ); Probably we should also use it for important i386 fields, to ensure they are not moved. #ifdef __i386__ C_ASSERT( FIELD_OFFSET(TEB, SpareBytes1) + FIELD_OFFSET(struct ntdll_thread_data, vm86) == FIELD_OFFSET(TEB, GdiTebBatch) ); C_ASSERT( FIELD_OFFSET(TEB, SpareBytes1) + FIELD_OFFSET(struct ntdll_thread_data, vm86) == 0x1fc ); C_ASSERT( FIELD_OFFSET(TEB, SpareBytes1) + FIELD_OFFSET(struct ntdll_thread_data, gs) == 0x1d8 ); #endif > diff --git a/dlls/ntdll/signal_x86_64.c b/dlls/ntdll/signal_x86_64.c > index f33fe4c..d2b040e 100644 > --- a/dlls/ntdll/signal_x86_64.c > +++ b/dlls/ntdll/signal_x86_64.c > @@ -1649,7 +1649,9 @@ static inline BOOL is_inside_signal_stack( void *ptr ) > */ > static void save_context( CONTEXT *context, const ucontext_t *sigcontext ) > { > - context->ContextFlags = CONTEXT_CONTROL | CONTEXT_INTEGER | CONTEXT_SEGMENTS; > + struct ntdll_thread_data * const regs = ntdll_get_thread_data(); > + > + context->ContextFlags = CONTEXT_CONTROL | CONTEXT_INTEGER | CONTEXT_SEGMENTS | CONTEXT_DEBUG_REGISTERS; > context->Rax = RAX_sig(sigcontext); > context->Rcx = RCX_sig(sigcontext); > context->Rdx = RDX_sig(sigcontext); > @@ -1686,6 +1688,12 @@ static void save_context( CONTEXT *context, const ucontext_t *sigcontext ) > #else > __asm__("movw %%ss,%0" : "=m" (context->SegSs)); > #endif > + context->Dr0 = regs->dr0; > + context->Dr1 = regs->dr1; > + context->Dr2 = regs->dr2; > + context->Dr3 = regs->dr3; > + context->Dr6 = regs->dr6; > + context->Dr7 = regs->dr7; > if (FPU_sig(sigcontext)) > { > context->ContextFlags |= CONTEXT_FLOATING_POINT; > @@ -1702,6 +1710,14 @@ static void save_context( CONTEXT *context, const ucontext_t *sigcontext ) > */ > static void restore_context( const CONTEXT *context, ucontext_t *sigcontext ) > { > + struct ntdll_thread_data * const regs = ntdll_get_thread_data(); > + > + regs->dr0 = context->Dr0; > + regs->dr1 = context->Dr1; > + regs->dr2 = context->Dr2; > + regs->dr3 = context->Dr3; > + regs->dr6 = context->Dr6; > + regs->dr7 = context->Dr7; > RAX_sig(sigcontext) = context->Rax; > RCX_sig(sigcontext) = context->Rcx; > RDX_sig(sigcontext) = context->Rdx; > @@ -1861,6 +1877,16 @@ __ASM_GLOBAL_FUNC( set_full_cpu_context, > void set_cpu_context( const CONTEXT *context ) > { > DWORD flags = context->ContextFlags & ~CONTEXT_AMD64; > + > + if (flags & CONTEXT_DEBUG_REGISTERS) > + { > + ntdll_get_thread_data()->dr0 = context->Dr0; > + ntdll_get_thread_data()->dr1 = context->Dr1; > + ntdll_get_thread_data()->dr2 = context->Dr2; > + ntdll_get_thread_data()->dr3 = context->Dr3; > + ntdll_get_thread_data()->dr6 = context->Dr6; > + ntdll_get_thread_data()->dr7 = context->Dr7; > + } > if (flags & CONTEXT_FULL) > { > if (!(flags & CONTEXT_CONTROL)) > @@ -2559,7 +2585,7 @@ static void raise_segv_exception( EXCEPTION_RECORD *rec, CONTEXT *context ) > } > break; > } > - status = raise_exception( rec, context, TRUE ); > + status = NtRaiseException( rec, context, TRUE ); > if (status) raise_status( status, rec ); NtRaiseException will always return a status != 0, so there is no need for the check anymore. > done: > set_cpu_context( context ); > @@ -2567,13 +2593,43 @@ done: > > > /********************************************************************** > + * raise_trap_exception > + */ > +static void raise_trap_exception( EXCEPTION_RECORD *rec, CONTEXT *context ) > +{ > + NTSTATUS status; > + > + if (rec->ExceptionCode == EXCEPTION_SINGLE_STEP) > + { > + /* when single stepping can't tell whether this is a hw bp or a > + * single step interrupt. try to avoid as much overhead as possible > + * and only do a server call if there is any hw bp enabled. */ > + > + if( !(context->EFlags & 0x100) || (ntdll_get_thread_data()->dr7 & 0xff) ) > + { > + /* (possible) hardware breakpoint, fetch the debug registers */ > + DWORD saved_flags = context->ContextFlags; > + context->ContextFlags = CONTEXT_DEBUG_REGISTERS; > + NtGetContextThread(GetCurrentThread(), context); > + context->ContextFlags |= saved_flags; /* restore flags */ > + } > + > + context->EFlags &= ~0x100; /* clear single-step flag */ > + } > + > + status = NtRaiseException( rec, context, TRUE ); > + raise_status( status, rec ); > +} > + > + > +/********************************************************************** > * raise_generic_exception > * > * Generic raise function for exceptions that don't need special treatment. > */ > static void raise_generic_exception( EXCEPTION_RECORD *rec, CONTEXT *context ) > { > - NTSTATUS status = raise_exception( rec, context, TRUE ); > + NTSTATUS status = NtRaiseException( rec, context, TRUE ); > if (status) raise_status( status, rec ); > set_cpu_context( context ); Similar to above, this can be simplified because status is always != 0. > } > @@ -2684,7 +2740,7 @@ static void segv_handler( int signal, siginfo_t *siginfo, void *sigcontext ) > */ > static void trap_handler( int signal, siginfo_t *siginfo, void *sigcontext ) > { > - EXCEPTION_RECORD *rec = setup_exception( sigcontext, raise_generic_exception ); > + EXCEPTION_RECORD *rec = setup_exception( sigcontext, raise_trap_exception ); > > switch (siginfo->si_code) > { > diff --git a/dlls/ntdll/thread.c b/dlls/ntdll/thread.c > index c9a5da6..934ada1 100644 > --- a/dlls/ntdll/thread.c > +++ b/dlls/ntdll/thread.c > @@ -744,10 +744,14 @@ NTSTATUS WINAPI NtSetContextThread( HANDLE handle, const CONTEXT *context ) > DWORD dummy, i; > BOOL self; > > -#ifdef __i386__ > +#if defined(__i386__) || defined(__x86_64__) > /* on i386 debug registers always require a server call */ > self = (handle == GetCurrentThread()); > +#ifdef __i386__ > if (self && (context->ContextFlags & (CONTEXT_DEBUG_REGISTERS & ~CONTEXT_i386))) > +#elif defined(__x86_64__) > + if (self && (context->ContextFlags & (CONTEXT_DEBUG_REGISTERS & ~CONTEXT_AMD64))) > +#endif > { > self = (ntdll_get_thread_data()->dr0 == context->Dr0 && > ntdll_get_thread_data()->dr1 == context->Dr1 && > @@ -903,9 +907,13 @@ NTSTATUS WINAPI NtGetContextThread( HANDLE handle, CONTEXT *context ) > copy_context( context, &ctx, ctx.ContextFlags & needed_flags ); > context->ContextFlags |= ctx.ContextFlags & needed_flags; > } > -#ifdef __i386__ > +#if defined(__i386__) || defined(__x86_64__) > /* update the cached version of the debug registers */ > +#ifdef __i386__ > if (context->ContextFlags & (CONTEXT_DEBUG_REGISTERS & ~CONTEXT_i386)) > +#elif defined(__x86_64__) > + if (context->ContextFlags & (CONTEXT_DEBUG_REGISTERS & ~CONTEXT_AMD64)) > +#endif > { > ntdll_get_thread_data()->dr0 = context->Dr0; > ntdll_get_thread_data()->dr1 = context->Dr1; > -- 2.7.4 >

2 2