January 2014 - Wine-devel - list.winehq.org

Better managing icons files names
by Pedro Henrique Lara Campos 02 Jan '14

02 Jan '14

Can Wine or could it in newer versions, use a prefix for their icons, like wine_56BA_Tcharam.png, or maybe their own folder? My ~/.local/share/icons is like... full of bottles, every corner I look, I find an icon from wine :P

3 2

Re: [PATCH] msxml3: avoid double free of str (Coverity)
by Marvin 01 Jan '14

01 Jan '14

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://newtestbot.winehq.org/JobDetails.pl?Key=3919 Your paranoid android. === wxppro (32 bit domdoc) === domdoc.c:9860: Test failed: got 00244F94

1 0

Re: [PATCH 12/21] kernel32: conditional variable support (try 6)
by Marcus Meissner 01 Jan '14

01 Jan '14

Hmm, THis should have been patch 0021 instead of 0012 , please disregard. Ciao, Marcus On Wed, Jan 01, 2014 at 07:44:16PM +0100, Marcus Meissner wrote: > Hi, > > Additional tightening and cleaning of comments. > > After Alexandres last guidance ("no global allocation") > and some hours of hair pulling I now did a implementation > with just a linked list of waiters. > > variable->Ptr points to the head of this list. > > The bit 0 of variable->Ptr is used for atomic locking of > the list and also its content, avoiding need for a lock variable. > > Small issues: > We are busy waiting on the bit 0 lock, not the most performance > wise best thing. > > Ciao, Marcus > --- > dlls/kernel32/kernel32.spec | 4 ++ > dlls/kernel32/sync.c | 151 +++++++++++++++++++++++++++++++++++++++++++ > 2 files changed, 155 insertions(+) > > diff --git a/dlls/kernel32/kernel32.spec b/dlls/kernel32/kernel32.spec > index 7779a75..e8d2bf0 100644 > --- a/dlls/kernel32/kernel32.spec > +++ b/dlls/kernel32/kernel32.spec > @@ -750,6 +750,7 @@ > @ stdcall InitOnceComplete(ptr long ptr) > @ stdcall InitOnceExecuteOnce(ptr ptr ptr ptr) > @ stdcall InitOnceInitialize(ptr) ntdll.RtlRunOnceInitialize > +@ stdcall InitializeConditionVariable(ptr) > @ stdcall InitializeCriticalSection(ptr) > @ stdcall InitializeCriticalSectionAndSpinCount(ptr long) > @ stdcall InitializeCriticalSectionEx(ptr long long) > @@ -1194,6 +1195,7 @@ > @ stdcall SignalObjectAndWait(long long long long) > @ stdcall SizeofResource(long long) > @ stdcall Sleep(long) > +@ stdcall SleepConditionVariableCS(ptr ptr long) > @ stdcall SleepEx(long long) > @ stdcall SuspendThread(long) > @ stdcall SwitchToFiber(ptr) > @@ -1263,6 +1265,8 @@ > @ stdcall WaitForSingleObjectEx(long long long) > @ stdcall WaitNamedPipeA (str long) > @ stdcall WaitNamedPipeW (wstr long) > +@ stdcall WakeAllConditionVariable (ptr) > +@ stdcall WakeConditionVariable (ptr) > @ stdcall WerRegisterFile(wstr long long) > @ stdcall WerRegisterMemoryBlock(ptr long) > @ stdcall WerRegisterRuntimeExceptionModule(wstr ptr) > diff --git a/dlls/kernel32/sync.c b/dlls/kernel32/sync.c > index 3c9e42a..34f39cd 100644 > --- a/dlls/kernel32/sync.c > +++ b/dlls/kernel32/sync.c > @@ -2327,3 +2327,154 @@ __ASM_STDCALL_FUNC(InterlockedDecrement, 4, > "ret $4") > > #endif /* __i386__ */ > + > +/* See http://locklessinc.com/articles/mutex_cv_futex/ for a futex implementation > + * See http://www.cs.wustl.edu/~schmidt/win32-cv-1.html for various implementations > + * of pthread_cond_* with Win32 primitives. > + * See glibc/nptl/DESIGN-condvar.txt for the glibc implementation > + * > + * This implementation uses a linked list as waiter queue, and the list > + * and its content is locked for access via bit 0 of variable->Ptr. > + * The actual waiting is done using a Win32 Event. > + */ > + > +struct _condvar_intern { > + struct _condvar_intern *next; > + HANDLE event; > +}; > + > +/********************************************************************** > + * InitializeConditionVariable (KERNEL32.@) > + */ > +VOID WINAPI InitializeConditionVariable(PCONDITION_VARIABLE variable) > +{ > + variable->Ptr = NULL; > +} > + > +/* > + * We lock the condition variable by setting bit 0 in variable->Ptr > + * atomically. We busy loop until we succeed and return the current value. > + * We also lock the empty list (NULL). > + * > + * List unlocking is done with "variable->Ptr = <retvalue>" > + */ > +static struct _condvar_intern* > +_lock_list (PCONDITION_VARIABLE variable) { > + struct _condvar_intern *oldvar, *var; > + > + do { > + __asm__ __volatile__("":::"memory"); > + var = variable->Ptr; > + if ((DWORD_PTR)var & 1) continue; /* locked already, retry. */ > + > + oldvar = InterlockedCompareExchangePointer (&variable->Ptr, (void*)((DWORD_PTR)var|1), var); > + if (oldvar == var) > + break; > + /* failed atomic locking, retry. */ > + } while (1); > + return var; > +} > + > +/********************************************************************** > + * WakeConditionVariable (KERNEL32.@) > + */ > +VOID WINAPI WakeConditionVariable(PCONDITION_VARIABLE variable) > +{ > + struct _condvar_intern *var; > + > + if (!variable->Ptr) > + return; > + > + var = _lock_list (variable); > + if (!var) { > + variable->Ptr = NULL; > + return; > + } > + > + /* remove and wake up the first entry */ > + NtSetEvent (var->event, NULL); > + > + variable->Ptr = var->next; > +} > + > +/********************************************************************** > + * WakeAllConditionVariable (KERNEL32.@) > + */ > +VOID WINAPI WakeAllConditionVariable(PCONDITION_VARIABLE variable) > +{ > + struct _condvar_intern *var; > + > + if (!variable->Ptr) > + return; > + > + var = _lock_list (variable); > + > + /* dequeue the whole waiter chain and wake it up */ > + while (var) { > + NtSetEvent (var->event, NULL); > + var = var->next; > + } > + > + variable->Ptr = NULL; > +} > + > +/********************************************************************** > + * SleepConditionVariableCS (KERNEL32.@) > + */ > +BOOL WINAPI SleepConditionVariableCS(PCONDITION_VARIABLE variable, > + LPCRITICAL_SECTION crit, DWORD timeout > +) > +{ > + struct _condvar_intern myself; > + struct _condvar_intern **xvar, *var; > + BOOL ret = TRUE; > + OBJECT_ATTRIBUTES attr; > + DWORD res; > + > + attr.Length = sizeof(attr); > + attr.RootDirectory = 0; > + attr.Attributes = OBJ_INHERIT; > + attr.ObjectName = NULL; > + attr.SecurityDescriptor = NULL; > + attr.SecurityQualityOfService = NULL; > + > + NtCreateEvent (&myself.event, EVENT_ALL_ACCESS, &attr, NotificationEvent, 0); > + myself.next = NULL; > + > + var = _lock_list (variable); > + /* Attach our stack based struct to the end of the list. > + * (at the end for fairness) > + */ > + xvar = &var; > + while (*xvar) > + xvar = &(*xvar)->next; > + *xvar = &myself; > + variable->Ptr = var; > + > + RtlLeaveCriticalSection(crit); > + > + res = WaitForSingleObject(myself.event, timeout); > + > + var = _lock_list (variable); > + > + /* Remove our entry */ > + xvar = (struct _condvar_intern **)&var; > + while (*xvar) { > + if (*xvar == &myself) { > + *xvar = (*xvar)->next; > + break; > + } > + xvar = &(*xvar)->next; > + } > + variable->Ptr = var; > + > + CloseHandle (myself.event); > + RtlEnterCriticalSection(crit); > + > + if (res == WAIT_TIMEOUT) { > + ret = FALSE; > + /* expected from caller, but not set by WaitForSingleObject */ > + SetLastError(ERROR_TIMEOUT); > + } > + return ret; > +} > -- > 1.7.10.4 > > > >

1 0

Re: [PATCH] msxml3: avoid double free of str (Coverity)
by Nikolay Sivov 01 Jan '14

01 Jan '14

On 1/1/2014 22:15, Marcus Meissner wrote: > 1127305 Double free > --- > dlls/msxml3/tests/domdoc.c | 1 - > 1 file changed, 1 deletion(-) > > diff --git a/dlls/msxml3/tests/domdoc.c b/dlls/msxml3/tests/domdoc.c > index df11197..f653a8b 100644 > --- a/dlls/msxml3/tests/domdoc.c > +++ b/dlls/msxml3/tests/domdoc.c > @@ -11047,7 +11047,6 @@ static void test_xmlns_attribute(void) > ok( hr == S_OK, "got 0x%08x\n", hr); > todo_wine ok( lstrcmpW(str, _bstr_("<Testing xmlns:dt=\"urn:schemas-microsoft-com:datatypes\" dt:dt=\"bin.base64\"/>")) == 0, > "got %s\n", wine_dbgstr_w(str)); > - SysFreeString(str); This is a correct fix technically, but it's better to remove second SysFreeString(), not this one. Speaking of style, I'd probably remove all 'if (hr == S_OK)' check, it never fails, or reduced a scope of 'str'.

1 0

Re: [PATCH] msxml3: avoid double free of str (Coverity)
by Marvin 01 Jan '14

01 Jan '14

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://newtestbot.winehq.org/JobDetails.pl?Key=3918 Your paranoid android. === wxppro (32 bit domdoc) === domdoc.c:9860: Test failed: got 00244F94

1 0

about eventlog
by Akira Nakagawa 01 Jan '14

01 Jan '14

I examined the design of RegistereventSource(). The return value type is HANDLE,but the details of it is unknown. However,I suppose,the code should be like this(I'm not good at coding of Win32) /****************************************************************************** * RegisterEventSourceW [ADVAPI32.@] * * See RegisterEventSourceA. */ HANDLE WINAPI RegisterEventSourceW( LPCWSTR lpUNCServerName, LPCWSTR lpSourceName ) { int arg_length,path_length; LPDWORD existance; LPWSTR reg_fullpath; LPWSTR filename; LPDWORD lptype; LPWSTR lpdata,gldata; HKEY hkeyp; HANDLE eventhandle; int number_of_entry,length_of_dat; FIXME("(%s,%s): stub\n", debugstr_w(lpUNCServerName), debugstr_w(lpSourceName)); if(lpUNCServerName!=NULL) { FIXME("lpUNCServerName:%s\nExecution in another server is not suppported.", debugstr_w(lpUNCServerName)); } else { existance = NULL; number_of_entry = 0; length_of_dat = 0; filename = (LPWSTR)"Application";//FIXME:this should be all of sub keies. reg_fullpath = lstrcatW((LPWSTR)"System\\CurrentControlSet\\Sevices\\Eventlog\\Application\\",lpSourceName); arg_length = lstrlenW(lpSourceName); RegCreateKeyExW(HKEY_LOCAL_MACHINE,reg_fullpath,0, NULL,REG_OPTION_NON_VOLATILE,KEY_READ,NULL,&hkeyp,existance);//Create or open key. RegQueryInfoKeyW(hkeyp,NULL,0,NULL,NULL,NULL,NULL, (LPDWORD)number_of_entry,NULL,(LPDWORD)length_of_dat,NULL,NULL); if(length_of_dat<2 || number_of_entry<2)//Key entry is void. return NULL; *lptype=(REG_SZ); RegQueryValueExW(hkeyp,(LPWSTR)"EventMessageFile",NULL,lptype,(LPBYTE)lpdata,(LPDWORD)path_length); RegCloseKey(hkeyp); eventhandle = GlobalAlloc(GHND,(arg_length+path_length)*sizeof(LPWSTR)); gldata=(LPWSTR)GlobalLock(eventhandle); //After that,store data follows:filename,file path,source name;I do not know how to. if(eventhandle==INVALID_HANDLE_VALUE) { ERR_(eventlog)("%s\n", debugstr_w(lpUNCServerName)); return NULL; } } return eventhandle; }

1 0

Re: [PATCH 3/3] ws2_32: Add an interactive test for interface change notifications (rebased).
by Marvin 01 Jan '14

01 Jan '14

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://newtestbot.winehq.org/JobDetails.pl?Key=3915 Your paranoid android. === wxppro (32 bit sock) === sock.c:554: Test failed: oob_server (5a8): unexpectedly at the OOB mark: 0 sock: unhandled exception c0000005 at 71AB4047 === w2008s64 (32 bit sock) === sock.c:554: Test failed: oob_server (750): unexpectedly at the OOB mark: 0 sock: unhandled exception c0000005 at 00000000 === w2008s64 (64 bit sock) === sock.c:554: Test failed: oob_server (654): unexpectedly at the OOB mark: 0 sock: unhandled exception c0000005 at 000007FEFE4C9CD9

1 0