[Bug 41713] New: Explorer.exe is wrong when started from the process creating the Wine prefix

https://bugs.winehq.org/show_bug.cgi?id=41713

Bug ID: 41713
           Summary: Explorer.exe is wrong when started from the process
                    creating the Wine prefix
           Product: Wine
           Version: 1.9.8
          Hardware: x86
                OS: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: -unknown
          Assignee: [e-mail]
          Reporter: [e-mail]
      Distribution: ---

To reproduce this do the following:

rm -rf ~/.wine
    cd dlls/shell32/tests
    rm shelldispatch.ok
    make shelldispatch.ok

This normally results in the following failure:

shelldispatch.c:624: Test failed: got 0x80040152

Which corresponds to the following line:
    ok(hr == S_OK, "got 0x%08x\n", hr);
    /* TODO: remove when explorer startup with clean prefix is fixed */
    if (hr != S_OK)
        return;

The workaround is to ensure the Wine prefix has been created before running the
test:

rm -rf ~/.wine
    ./wine hostname
    ./server/wineserver -w
    cd dlls/shell32/tests
    rm shelldispatch.ok
    make shelldispatch.ok

Note that wt-daily is now implementing this workaround (see
https://github.com/fgouget/wt-daily).

[Bug 41713] Explorer.exe is wrong when started from the process creating the Wine prefix

https://bugs.winehq.org/show_bug.cgi?id=41713

Zebediah Figura  changed:

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

--- Comment #1 from Zebediah Figura  ---
So the problem here is that the ShellWindows class belongs to explorer.exe, and
must be registered as local server by explorer.exe.

wineboot does the following things in order:

* create a dialog window notifying the user that Wine is updating its prefix
* update the prefix, with rundll32.exe
* start the application

Both creating this dialog window, and using rundll32.exe (for reasons I am
unsure of) start explorer.exe, which attempts to register the ShellWindows
class. This fails, because in order to do this we need the marshalling classes,
and those haven't been registered yet. This is what's responsible for the error
messages that get printed on every startup:

err:ole:marshal_object couldn't get IPSFactory buffer for interface
{00000131-0000-0000-c000-000000000046}
err:ole:marshal_object couldn't get IPSFactory buffer for interface
{6d5140c1-7436-11ce-8034-00aa006009fa}
err:ole:StdMarshalImpl_MarshalInterface Failed to create ifstub,
hres=0x80004002
err:ole:CoMarshalInterface Failed to marshal the interface
{6d5140c1-7436-11ce-8034-00aa006009fa}, 80004002
err:ole:get_local_server_stream Failed: 80004002
err:ole:marshal_object couldn't get IPSFactory buffer for interface
{00000131-0000-0000-c000-000000000046}
err:ole:marshal_object couldn't get IPSFactory buffer for interface
{6d5140c1-7436-11ce-8034-00aa006009fa}
err:ole:StdMarshalImpl_MarshalInterface Failed to create ifstub,
hres=0x80004002
err:ole:CoMarshalInterface Failed to marshal the interface
{6d5140c1-7436-11ce-8034-00aa006009fa}, 80004002
err:ole:get_local_server_stream Failed: 80004002

The most obvious solution, as it seems to me, is to start explorer.exe a third
time in wineboot.exe, to ensure that those classes get registered.

[Bug 41713] Explorer.exe is wrong when started from the process creating the Wine prefix

https://bugs.winehq.org/show_bug.cgi?id=41713

--- Comment #2 from Nikolay Sivov  ---
The actual issue I think was that registry is not fully populated by the time
we try to access interface/proxy keys.

[Bug 41713] Explorer.exe is wrong when started from the process creating the Wine prefix

https://bugs.winehq.org/show_bug.cgi?id=41713

--- Comment #3 from Zebediah Figura  ---
(In reply to Nikolay Sivov from comment #2)
> The actual issue I think was that registry is not fully populated by the
> time we try to access interface/proxy keys.

Right, isn't that what I said? Or am I wrong somehow?

[Bug 41713] Explorer.exe is wrong when started from the process creating the Wine prefix

https://bugs.winehq.org/show_bug.cgi?id=41713

--- Comment #4 from Nikolay Sivov  ---
(In reply to Zebediah Figura from comment #3)
> (In reply to Nikolay Sivov from comment #2)
> > The actual issue I think was that registry is not fully populated by the
> > time we try to access interface/proxy keys.
>
> Right, isn't that what I said? Or am I wrong somehow?

Yes, sorry, I guess I skipped important part when reading your comment.

[Bug 41713] Explorer.exe is wrong when started from the process creating the Wine prefix

https://bugs.winehq.org/show_bug.cgi?id=41713

Fabian Maurer  changed:

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

--- Comment #5 from Fabian Maurer  ---
*** Bug 45232 has been marked as a duplicate of this bug. ***

[Bug 41713] Explorer.exe is wrong when started from the process creating the Wine prefix

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=41713

--- Comment #6 from Anastasius Focht  ---
*** Bug 45600 has been marked as a duplicate of this bug. ***

[Bug 41713] Initial 'explorer' process fails to register/run ShellWindows class local COM server in new WINEPREFIX due to incomplete registry data

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=41713

Anastasius Focht  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|Explorer.exe is wrong when  |Initial 'explorer' process
                   |started from the process    |fails to register/run
                   |creating the Wine prefix    |ShellWindows class local
                   |                            |COM server in new
                   |                            |WINEPREFIX due to
                   |                            |incomplete registry data
                 CC|                            |[e-mail]
          Component|-unknown                    |programs

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

revisiting, still present. I've encountered this in bug 45600 again, see
https://bugs.winehq.org/show_bug.cgi?id=45600#c3

If you don't mind I refine the summary line a bit more. It's not easy to
find/remember in Bugzilla. I suspect a number of apps suffering from this and
users currently just get away by chance:

* WINEPREFIX recycling (using existing prefix)
* manual bootstrapping/creation of WINEPREFIX
* unelevated process starts using "ShellExecuteFromExplorer" method happen
separately, after installer has terminated (launcher)

How about propagating the HRESULT from 'shellwindows_init' to 'manage_desktop'
caller and let the process terminate if the COM server fails to register/run
due to incomplete registry data? This also ensures that the "root" instance
which can't run a COM server by design (only pumps messages) is gone.
Then either let the next (UI) process that needs the desktop window
automatically bootstrap 'explorer.exe' process again or do it explicitly in
'wineboot' at late stage. At that point all the proxy data is in registry.

https://source.winehq.org/git/wine.git/blob/HEAD:/programs/explorer/desktop....

https://source.winehq.org/git/wine.git/blob/HEAD:/programs/explorer/desktop....

https://source.winehq.org/git/wine.git/blob/HEAD:/programs/explorer/desktop....

$ wine --version
wine-5.0-144-g9a9a1821a3

Regards

[Bug 41713] Initial 'explorer' process fails to register/run ShellWindows class local COM server in new WINEPREFIX due to incomplete registry data

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=41713

--- Comment #8 from Anastasius Focht  ---
Hello again,

adding another way to reproduce the problem without even having tests compiled.

Taken from my analysis https://bugs.winehq.org/show_bug.cgi?id=45600#c3

--- snip ---
# run against existing WINEPREFIX
# 9ba05972-f6a8-11cf-a442-00a0c90a8f39 -> CLSID_ShellWindows

$ while true ; do wine cmd.exe /c "echo test >
\\.\pipe\{9ba05972-f6a8-11cf-a442-00a0c90a8f39}" ; sleep 1 ; done

0034:fixme:ntdll:server_get_file_info Unsupported info class 14
Invalid function.

0036:fixme:ntdll:server_get_file_info Unsupported info class 14
Invalid function.

...
# kill explorer process/local COM server
...
File not found.
--- snip ---

'Invalid function.' -> ok, named pipe can be opened
'File not found.' -> error, named pipe does not exist (COM server not running)

If this is run against a newly created WINEPREFIX it will always fail. Just use
some builtin app to bootstrap and keep it from being shut down.

Regards

[Bug 41713] Initial 'explorer' process fails to register/run ShellWindows class local COM server in new WINEPREFIX due to incomplete registry data

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=41713

--- Comment #9 from Zebediah Figura  ---
Just to add another interesting variable to this, I have a set of patches which
actually uses IShellWindows for its intended purpose—that is, to locate
existing Explorer windows browsing a given path so that they can be brought to
foreground instead of duplicated. This naturally fails with the first window,
because explorer.exe isn't started until a window is created. I don't have a
good solution for this yet.

[Bug 41713] Initial 'explorer' process fails to register/run ShellWindows class local COM server in new WINEPREFIX due to incomplete registry data

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=41713

Nikolay Sivov  changed:

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

--- Comment #10 from Nikolay Sivov  ---
*** Bug 49918 has been marked as a duplicate of this bug. ***

[Bug 41713] Initial 'explorer' process fails to register/run ShellWindows class local COM server in new WINEPREFIX due to incomplete registry data

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=41713

mirh  changed:

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

[Bug 41713] Initial 'explorer' process fails to register/run ShellWindows class local COM server in new WINEPREFIX due to incomplete registry data

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=41713

--- Comment #11 from François Gouget  ---
This bug is still present in Wine 6.1 (433b9081ba7c).

[Bug 41713] Initial 'explorer' process fails to register/run ShellWindows class local COM server in new WINEPREFIX due to incomplete registry data

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=41713

Anastasius Focht  changed:

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

--- Comment #12 from Anastasius Focht  ---
*** Bug 50327 has been marked as a duplicate of this bug. ***