[Bug 35884] New: Visual Studio 20XX frequently calls CoGetCallerTID(), resulting in console spam due to stub

https://bugs.winehq.org/show_bug.cgi?id=35884

Bug ID: 35884
           Summary: Visual Studio 20XX frequently calls CoGetCallerTID(),
                    resulting in console spam due to stub
           Product: Wine
           Version: 1.7.15
          Hardware: x86
                OS: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: ole32
          Assignee: [e-mail]
          Reporter: [e-mail]

Hello folks,

as the summary says... not really harmful though.

--- snip ---
...
fixme:ole:CoGetCallerTID stub!
fixme:ole:CoGetCallerTID stub!
fixme:ole:CoGetCallerTID stub!
fixme:ole:CoGetCallerTID stub!
...
--- snip ---

I'd prefer a real implementation over a 'FIXME once', it's not particularly
hard to do.

MSDN:
http://msdn.microsoft.com/en-us/library/windows/desktop/ms680683%28v=vs.85%2...

--- quote ---

lpdwTID [out]

Receives the apartment ID of the caller's thread. For a single threaded
apartment (STA), this is the current thread ID. For a multithreaded apartment
(MTA), the value is 0. For a neutral apartment (NA), the value is -1.
--- quote ---

Source:
http://source.winehq.org/git/wine.git/blob/39562961283ccb7c838c5650b11873e69...

--- snip ---
92 HRESULT WINAPI CoGetCallerTID(LPDWORD lpdwTID)
93 {
94     FIXME("stub!\n");
95     return E_NOTIMPL;
96 }
--- snip ---

$ wine --version
wine-1.7.15-112-g2aad5d7

Regards

[Bug 35884] Visual Studio 20XX frequently calls CoGetCallerTID(), resulting in console spam due to stub

https://bugs.winehq.org/show_bug.cgi?id=35884

Alistair Leslie-Hughes  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

[Bug 35884] Visual Studio 20XX frequently calls CoGetCallerTID(), resulting in console spam due to stub

https://bugs.winehq.org/show_bug.cgi?id=35884

--- Comment #1 from Austin English  ---
This is your friendly reminder that there has been no bug activity for over a
year. Is this still an issue in current (1.7.51 or newer) wine?

[Bug 35884] Visual Studio 20XX frequently calls CoGetCallerTID(), resulting in console spam due to stub

https://bugs.winehq.org/show_bug.cgi?id=35884

--- Comment #2 from Anastasius Focht  ---
Hello folks,

obviously still present.

https://source.winehq.org/git/wine.git/blob/HEAD:/dlls/ole32/ole2stubs.c#l90

$ wine --version
wine-1.7.51

Regards

[Bug 35884] Visual Studio 20XX frequently calls CoGetCallerTID(), resulting in console spam due to stub

https://bugs.winehq.org/show_bug.cgi?id=35884

--- Comment #3 from Anastasius Focht  ---
Hello folks,

revisiting, obviously still present.

$ wine --version
wine-4.3

Regards

[Bug 35884] Visual Studio 20XX frequently calls CoGetCallerTID(), resulting in console spam due to stub

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=35884

--- Comment #4 from Nikolay Sivov  ---
This does not look trivial in case of remote calls, I don't see how client tid
could be communicated to the server.

[Bug 35884] Visual Studio 20XX frequently calls CoGetCallerTID(), resulting in console spam due to stub

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=35884

Anastasius Focht  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |download
                URL|                            |https://download.microsoft.
                   |                            |com/download/A/9/1/A91D6B2B
                   |                            |-A798-47DF-9C7E-A97854B7DD1
                   |                            |8/VC.iso

--- Comment #5 from Anastasius Focht  ---
Hello Nikolay,

--- quote ---
This does not look trivial in case of remote calls, I don't see how client tid
could be communicated to the server.
--- quote ---

I don't know either, but I found this interesting piece of information about
the location of COM caller/callee process and thread ids:

https://stackoverflow.com/questions/41204672/problems-using-dbgrpc-on-window...
("Problems using dbgrpc on Windows7"):

--- quote ---
To avoid this kernel mess, I use the "ReservedForOle" pointer in the thread's
TEB (offset 0xf80) where COM stores process and thread IDs. The following
commands can be used to access them for WinDbg:

In COM-server: where is an incoming COM-call is coming from: Caller's process
ID: ? dwo(dwo(@$teb + 0xf80) + 0x108) Caller's thread ID (will be 0 if the
caller's thread is in MTA, or -1 if in NA): ? dwo(dwo(@$teb + 0xf80) + 0x34)

In COM-client: where is an outgoing COM-call going to: Target's process ID: ?
dwo(dwo(@$teb + 0xf80) + 0x100) Target's thread ID (will be 0 if the target
server is a MTA COM-server): ? dwo(dwo(@$teb + 0xf80) + 0x104)

These values are for 32-bit processes. For native 64-bit processes, the offsets
would be different (e.g. "ReservedForOle" is on the offset 0x1758 in TEB).
--- quote ---

At least we have explanation for the 'unknown' DWORD at offset +0x34 and more
offsets.

https://source.winehq.org/git/wine.git/blob/3b1c3e57fcac4e23c0f9b376a483c494...

--- snip ---
 163 /* this is what is stored in TEB->ReservedForOle */
 164 struct oletls
 165 {
 166     struct apartment *apt;
 167     IErrorInfo       *errorinfo;   /* see errorinfo.c */
 168     DWORD             thread_seqid;/* returned with CoGetCurrentProcess */
 169     DWORD             apt_mask;    /* apartment mask (+0Ch on x86) */
 170     void            *unknown0;
 171     DWORD            inits;        /* number of times CoInitializeEx
called */
 172     DWORD            ole_inits;    /* number of times OleInitialize called
*/
 173     GUID             causality_id; /* unique identifier for each COM call
*/
 174     LONG             pending_call_count_client; /* number of client calls
pending */
 175     LONG             pending_call_count_server; /* number of server calls
pending */
 176     DWORD            unknown;
 177     IObjContext     *context_token; /* (+38h on x86) */
 178     IUnknown        *call_state;    /* current call context (+3Ch on x86)
*/
 179     DWORD            unknown2[46];
 180     IUnknown        *cancel_object; /* cancel object set by
CoSetCancelObject (+F8h on x86) */
 181     IUnknown        *state;       /* see CoSetState */
 182     struct list      spies;         /* Spies installed with
CoRegisterInitializeSpy */
 183     DWORD            spies_lock;
 184 };
--- snip ---

Also adding download links.

https://download.microsoft.com/download/A/9/1/A91D6B2B-A798-47DF-9C7E-A97854...
(MS Visual C++ 2005 Express)

Internet Archive Snapshot/Mirror:

https://web.archive.org/web/20190127082540/http://download.microsoft.com/dow...

$ sha1sum VC.iso 
1ae44e4eaf8c61c3a39e573fd6efd9889e940529  VC.iso

$ du -sh VC.iso 
463M    VC.iso

$ wine --version
wine-5.0-rc3

Regards

[Bug 35884] Visual Studio 20XX frequently calls CoGetCallerTID(), resulting in console spam due to stub

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=35884

--- Comment #6 from Nikolay Sivov  ---
Improved layout with more fields matching on x64 would be something like this:

---
struct oletls
{
    struct apartment *apt;
    IErrorInfo       *errorinfo;   /* see errorinfo.c */
    DWORD             thread_seqid;/* (+8h/+10h) returned with
CoGetCurrentProcess */
    DWORD             apt_mask;    /* (+Ch/+14h) apartment mask */
    DWORD             inits;
    DWORD_PTR         unknown1;
    DWORD             ole_inits;
    DWORD             unknown2;
    DWORD_PTR         unknown3[5];
    DWORD             caller_tid;    /* (+34h/+58h) CoGetCallerTID() */
    DWORD_PTR         unknown4;
    IObjContext      *context_token; /* (+38h/+68h on x86) */
    IUnknown         *call_state;    /* current call context (+3Ch on x86) */
    DWORD             unknown5[12];
    GUID              causality_id;  /* (+80h/+d8h) */
    DWORD             unknown6[18];
    IUnknown         *cancel_object; /* cancel object set by CoSetCancelObject
(+F8h on x86) */
    IUnknown         *state;       /* see CoSetState */
    struct list       spies;         /* Spies installed with
CoRegisterInitializeSpy */
    DWORD             spies_lock;
    LONG              pending_call_count_client; /* number of client calls
pending */
    LONG              pending_call_count_server; /* number of server calls
pending */
};
---

So caller tid is at 0x58. Function seems to simply return this field value.

Regarding actually implementing this, I think it could about the same as
causality id being set/restored around RPC_ExecuteCall(), which real question
being how this TID is communicated through RPC_MESSAGE. (This is probably
limited to same machine case, or even for certain local protocols)

[Bug 35884] Visual Studio 20XX frequently calls CoGetCallerTID(), resulting in console spam due to stub

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=35884

Anastasius Focht  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                URL|https://download.microsoft. |https://web.archive.org/web
                   |com/download/A/9/1/A91D6B2B |/20190127082540/http://down
                   |-A798-47DF-9C7E-A97854B7DD1 |load.microsoft.com/download
                   |8/VC.iso                    |/A/9/1/A91D6B2B-A798-47DF-9
                   |                            |C7E-A97854B7DD18/VC.iso

--- Comment #7 from Anastasius Focht  ---
Hello folks,

revisiting, still present.

$ wine --version
wine-6.0

Regards