[Bug 48993] 64-bit Windows Management Framework 2.0 for Windows 2003 installer fails with error 'Invalid handle' (needs 'scecli.SceConfigureSystem' stub)

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=48993

Anastasius Focht <focht at gmx.net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |patch
            Summary|64-bit Powershell 2.0 for   |64-bit Windows Management
                   |Windows 2003 installer      |Framework 2.0 for Windows
                   |fails due to missing        |2003 installer fails with
                   |'scecli.dll'                |error 'Invalid handle'
                   |                            |(needs
                   |                            |'scecli.SceConfigureSystem'
                   |                            |stub)

--- Comment #3 from Anastasius Focht <focht at gmx.net> ---
Hello Leith,

thanks for the patch. I'm refining the summary to highlight it's that one stub
function needed which implies the addition of the stub dll as well.

--- quote ---
Is there a way to verify the number of arguments on the stack expected by
SceConfigureSystem? 
--- quote ---

There are a few techniques if you don't know the prototype when no
documentation/SDK headers exist. All of that requires that the general rules of
the Wine project w.r.t. reverse engineering are obeyed.

1) Analyse the caller of the API function (= application) and check how the
stack is set up. Most of the time the number and type of arguments can be
deduced fairly easily. This technique should only be used if the caller itself
is not subject to reverse engineering restrictions (not part of OS)

2) Write a test case that does argument fuzzing, i.e. calls the unknown API
function with varying number of arguments. You start with one argument (DWORD
type) and check for stack imbalance (STDCALL).

--- quote ---
Similarly is there a way to know if the remaining arguments should be treated
as pointers, or as integers?
--- quote ---

After figuring the out the number of arguments you take a look at the value
range. Specific types (pointers, handle, enums) have distinct ranges. For
example if you see a value that matches the return/out value of preceding win32
API calls, i.e. CreateFile(), HeapAlloc(), MultiByteToWideChar() you can deduce
the type and purpose of parameters this way. For structure layout/unknown
members it's more complicated. You use "magic" patterns to figure out which
fields/offsets have been written to and interpret the value ranges again.

I wouldn't bother too much here since only a stub is needed. If you look at
Wine source, there are multiple places where stubs exist where only the number
of arguments are known, gathered by above mentioned techniques.

--- snip ---
$ egrep -Hrni "(unk1|unknown1)"
--- snip ---

Regards

-- 
Do not reply to this email, post in Bugzilla using the
above URL to reply.
You are receiving this mail because:
You are watching all bug changes.