C++ class constructor can be inlined or not be inlined. However, I found a strange situation where only inline class constructor can avoid Visual Studio memory crash. The example is as follows:
dll.h
class _declspec(dllexport) Image
{
public:
Image();
virtual ~Image();
};
class _declspec(dllexport) Testimage:public Image
{
public:
Testimage();
virtual ~Testimage();
};
typedef std::auto_ptr<Testimage> TestimagePtr;
dll.cpp
#include "dll.h"
#include <assert.h>
Image::~Image()
{
std::cout<<"Image is being deleted."<<std::endl;
}
Image::Image()
{
}
Testimage::Testimage()
{
}
Testimage::~Testimage()
{
std::cout<<"Geoimage is being deleted."<<std::endl;
}
The dll library is compiled as a dynamic library, and it is statically linked to the C++ runtime library (Multi-threaded Debug (/MTd)). The executable program that runs the library is as follows:
int main()
{
TestimagePtr my_img(new Testimage());
return 0;
}
The executable program will invoke the dll library and it also statically links the runtime library. The problem I have is that when running the executable program the following error message appears:
However, when the class constructor in dll is inlined as the following codes show:
class _declspec(dllexport) Image
{
public:
Image();
virtual ~Image();
};
class _declspec(dllexport) Testimage:public Image
{
public:
Testimage()
{
}
virtual ~Testimage();
};
The crash will disappear. Could someone explain the reason behind? Thanks! By the way, I am using VC2010.
EDIT: The following situation also trigger the same crash
.
Situation 1
int main()
{
//TestimagePtr my_img(new Testimage());
Testimage *p_img;
p_img = new Testimage();
delete p_img;
return 0;
}
it is statically linked to the C++ runtime library (Multi-threaded Debug (/MTd)
This is a very problematic scenario in versions of Visual Studio prior to VS2012. The issue is that you have more than one version of the CRT loaded in your process. One used by your EXE, another used by the DLL. This can cause many subtle problems, and not so subtle problems like this crash.
The CRT has global state, stuff like errno and strtok() cannot work properly when that global state is updated by one copy of the CRT and read back by another copy. Relevant to your crash, a hidden global state variable is the heap that the CRT uses to allocate memory from. Functions like malloc() and ::operator new use that heap.
This goes wrong when objects are allocated by one copy of the CRT and released by another. The pointer that's passed to free() or ::operator delete belongs to the wrong heap. What happens next depends on your operating system. A silent memory leak in XP. In Vista and up, you program runs with the debug version of the memory manager enabled. Which triggers a breakpoint when you have a debugger attached to your process to tell you that there's a problem with the pointer. The dialog in your screenshot is the result. It isn't otherwise very clear to me how inlining the constructor could make a difference, the fundamental issue however is that your code invokes undefined behavior. Which has a knack for producing random outcomes.
There are two approaches available to solve this problem. The first one is the simple one, just build both your EXE and your DLL project with the /MD compile option instead. This selects the DLL version of the CRT. It is now shared by both modules and you'll only have a single copy of the CRT in your process. So there is no longer a problem with having one module allocating and another module releasing memory, the same heap is used.
This will work fine to solve your problem but can still become an issue later. A DLL tends to live a life of its own and may some day be used by another EXE that was built with a different version of the CRT. The CRT will now again not be shared since they'll use different versions of the DLL, invoking the exact same failure mode you are seeing today.
The only way to guarantee that this cannot happen is to design your DLL interface carefully. And ensure that there will never be a case where the DLL allocates memory that the client code needs to release. That requires giving up on a lot of C++ goodies. You for example can never write a function that returns a C++ object, like std::string. And you can never allow an exception to cross the module boundary. You are basically down to a C-style interface. Note how COM addresses this problem by using interface-based programming techniques and a class factory plus reference counting to solve the memory management problem.
VS2012 has a counter-measure against this problem, it has a CRT version that allocates from the default process heap. Which solves this particular problem, not otherwise a workaround for the global state issue for other runtime functions. And adds some new problems, a DLL compiled with /MT that gets unloaded that doesn't release all of its allocations now causes an unpluggable leak for example.
This is an ugly problem in C++, the language fundamentally misses an ABI specification that addresses problems like this. The notion of modules is entirely missing from the language specification. Being worked on today but not yet completed. Not simple to do, it is solved in other languages like Java and the .NET languages by specifying a virtual machine, providing a runtime environment where memory management is centralized. Not the kind of runtime environment that excites C++ programmers.
I tried to reproduce your problem in VC2010 and it doesn't crash. It works with a constructor inline or not. Your problem is probably not in what you write here.
Your project is too hard to open as it seams to have its file pathes set in absolute, probably because generated with CMake. (So the files are not found by the compiler).
The problem I see in your code is that you declare the exported classes with _declspec(dllexport) directly written.
You should have a #Define to do this, and the value should be _declspec(dllimport) when read from the exe compilation. Maybe the problem comes from that.
I created c++ dll (using mingw) from code I wrote on linux (gcc), but somehow have difficulties using it in VC++. The dll basically exposes just one class, I created pure virtual interface for it and also factory function which creates the object (the only export) which looks like this:
extern "C" __declspec(dllexport) DeviceDriverApi* GetX5Driver();
I added extern "C" to prevent name mangling, dllexport is replaced by dllimport in actual code where I want to use the dll, DeviceDriverApi is the pure virtual interface.
Now I wrote simple code in VC++ which just call the factory function and then just tries to delete the pointer. It compiles without any problems but when I try to run it I get access violation error. If I try to call any method of the object I get access violation again.
When I compile the same code in MinGW (gcc) and use the same library, it runs without any problems. So there must be something (hehe, I guess many differences actually :)) between how VC++ code uses the library and gcc code.
Any ideas what?
Cheers,
Tom
Edit:
The code is:
DeviceDriverApi* x5Driver = GetX5Driver();
if (x5Driver->isConnected())
Console::WriteLine(L"Hello World");
delete x5Driver;
It's crashing when I try to call the method and when I try to delete the pointer as well. The object is created correctly though (the first line). There are some debug outputs when the object is created and I can see them before I get the access violation error.
You're using one compiler (mingw) for the DLL, and another (VC++) for the calling code.
You're calling a 'C' function, but returning a pointer to a C++ Object.
That will never work, because VTable layouts are almost guranteed to be incompatible. And, the DLL and app are probably using different memory managers, so you're doing new() with one and delete() with the other. Again, it just won't work.
For this to work the two compilers need to both support a standard ABI (Application Binary Interface). I don't think such a thing exists for Windows.
The best option is to expose all you DLL object methods and properties via C functions (including one to delete the object). You can the re-wrap into a C++ object on the calling end.
The two different compilers may be using different calling conventions. Try putting _cdecl before the function name in both the client code and the DLL code and recompiling both.
More info on calling conventions here: http://en.wikipedia.org/wiki/X86_calling_conventions
EDIT: The question was updated with more detail and it looks likely the problem is what Adrien Plisson describes at the end of his answer. You're creating an object in one module and freeing it in another, which is wrong.
(1) I suspect a calling covnention problem as well, though the simple suggestion by Leo doesn't seem to have helped.
Is isConnected virtual? It is possible that MinGW and VC++ use different implementations for a VTable, in which case, well, tough luck.
Try to see how far you get with the debugger: does it crash at the call, or the return? Do you arrive at invalid code? (If you know to read assembly, that usually helps a lot with these problems.)
Alternatively, add trace statements to the various methods, to see how far you get.
(2) For a public DLL interface, never free memory in the caller that was allocated by a callee (or vice versa). The DLL likely runs with a completely different heap, so the pointer is not known.
If you want to rely on that behavior, you need to make sure:
Caller and Callee (i.e. DLL and main program, in your case) are compiled with the same version of the sam compiler
for all supported compilers, you have configured the compile options to ensure caller and callee use the same shared runtime library state.
So the best way is to change your API to:
extern "C" __declspec(dllexport) DeviceDriverApi* GetX5Driver();
extern "C" __declspec(dllexport) void FreeDeviceDriver(DeviceDriverApi* driver);
and, at caller site, wrap in some way (e.g. in a boost::intrusive_ptr).
try looking at the imported libraries from both your DLL and your client executable. (you can use the Dependency Viewer or dumpbin or any other tool you like). verify that both the DLL and the client code are using the same C++ runtime.
if it is not the case, you can indeed run into some issues since the way the memory is managed may be different between the 2, leading to a crash when freeing from one runtime a pointer allocated from another runtime.
if this is really your problem, try not destroying the pointer in your client executable, but rather declare and export a function in your DLL which will take care of destroying the pointer.
I'm programing on C++, I'm using Visual Studio 2008, Windows XP, and I have the following problem:
My application, that is a DLL that can be used from Python, loads an external dll, uses the required methods, and then unloads this external Dll.
It's working properly, but after more than 1000 cycles the method "LoadLibraryA" returns a NULL reference.
The main steps are:
HINSTANCE h = NULL;
h = LoadLibraryA(dllfile.c_str());
DWORD dw = GetLastError();
The error got is:
ERROR_DLL_INIT_FAILED
1114 (0x45A) A dynamic link library (DLL) initialization routine failed.
The Dll is unloaded by using the following:
FreeLibrary(mDLL);
mDLL = NULL;
Where mDLL is defined like this:
HINSTANCE mDLL;
First alternative tried:
Just load the Dll only once, and unloaded it when the application ends. This fix the problem but introduces a new one.
When the application ends, instead of first executing the DllMain method of my applicaion, wich unloads the external DLL, is executing first the DllMain method of the other Dll. This cause the following error because my application is trying to unload a Dll that was unload by itself previously.
"Unhandled exception at 0x04a00d07 (DllName.DLL) in Python.exe: 0xC0000005: Access violation reading location 0x0000006b".
Any suggestion will be welcomed.
Thanks in advance.
Regards.
Make sure that initialization code of the loaded/unloaded library doesn't leak memory. Many libraries expect to be loaded only once and not always clean up their resources properly.
E.g. in C++ file at the top level one can declare and initialize a variable like this:
AClass *a = new AClass(1,2,3);
The code would be executed when library is loaded automatically. Yet, now, it is impossible to free the hanging instance as library doesn't know precisely when/how it is going to be unloaded. In the case one can either replace "AClass *a" with "AClass a" or write your own DllMain for the library and free resources on DLL_PROCESS_DETACH.
If you have no control over the library's code, then it might make sense to create a cache of loaded libraries and simply never unload them. It is very hard to imagine that there would be unlimited number of libraries to overload such cache.
I have a native DLL that is a plug-in to a different application (one that I have essentially zero control of). Everything works just great until I link with an additional .lib file (links my DLL to another DLL named ABQSMABasCoreUtils.dll). This file contains some additional API from the parent application that I would like to utilize. I haven't even written any code to use any of the functions exported but just linking in this new DLL is causing problems. Specifically, I get the following error when I attempt to run the program:
The application failed to initialize properly (0xc0000025). Click on OK to terminate the application.
I believe I have read somewhere that this is typically due to a DllMain function returning FALSE. Also, the following message is written to the standard output:
ERROR: Memory allocation attempted before component initialization
I am almost 100% sure this error message is coming from the application and is not some type of Windows error.
Looking into this a little more (aka flailing around and flipping every switch I know of) I linked with /MAP turned on and found this in the resulting .map file:
0001:000af220 ??3#YAXPEAX#Z 00000001800b0220 f ABQSMABasCoreUtils_import:ABQSMABasCoreUtils.dll
0001:000af226 ??2#YAPEAX_K#Z 00000001800b0226 f ABQSMABasCoreUtils_import:ABQSMABasCoreUtils.dll
0001:000af22c ??_U#YAPEAX_K#Z 00000001800b022c f ABQSMABasCoreUtils_import:ABQSMABasCoreUtils.dll
0001:000af232 ??_V#YAXPEAX#Z 00000001800b0232 f ABQSMABasCoreUtils_import:ABQSMABasCoreUtils.dll
If I undecorate those names using "undname" they give the following (same order):
void __cdecl operator delete(void * __ptr64)
void * __ptr64 __cdecl operator new(unsigned __int64)
void * __ptr64 __cdecl operator new[](unsigned __int64)
void __cdecl operator delete[](void * __ptr64)
I am not sure I understand how anything from ABQSMABasCoreUtils.dll can exist within this .map file or why my DLL is even attempting to load ABQSMABasCoreUtils.dll if I don't have any code that references this DLL. Can anyone help me put this information together and find out why this isn't working? For what it's worth I have confirmed via "dumpbin" that the parent application imports ABQSMABasCoreUtils.dll, so it is being loaded no matter what. I have also tried delay loading this DLL in my DLL but that did not change the results.
EDIT
I have double checked and all files involved are 64 bit.
I just had exactly the same problem. This is an issue with the Abaqus API rather than with the loading of DLLS.
I think it is because the Abaqus API overrides the new and delete functions (as you seem to have noticed). If you call new or delete in your program before initializing the Abaqus API, such as by calling odb_initializeAPI(); then you get the
ERROR: Memory allocation attempted before component initialization
error message and the program crashes.
In my program, calling odb_initializeAPI(); before the first new resolved the problem.
Well, sure you'll reference the imports of that library. Hard to write a C++ program without using the new or delete operator. Dealing with 3rd party software that thinks it needs to override the CRT version of those operators is hard enough, impossible when it won't allow you to call them until it thinks the time is right. Abandon all hope or seek help from the vendor.
One of the possible reason of an error during loading of ABQSMABasCoreUtils.dll is that some dependency module (inclusive delayed load DLLs) could not be found. Use Dependency Walker (see http://www.dependencywalker.com/) to examine all dependencies of ABQSMABasCoreUtils.dll.
I have two suggestions:
Verify that you can load ABQSMABasCoreUtils.dll with respect of LoadLibrary. You don't need call any function from ABQSMABasCoreUtils.dll. Usage of LoadLibrary I don't see as the end solution. It' s only a diagnostic test. With the test you can verify either you have some general problem of loading ABQSMABasCoreUtils.dll in your program or you have some kind of process initialization problem.
If loading of ABQSMABasCoreUtils.dll with respect of LoadLibrary will failed, then use profiling feature of Dependency Walker to protocol of all calls done during loading of ABQSMABasCoreUtils.dll. One other way would be usage of Process Monitor (see http://technet.microsoft.com/en-us/sysinternals/bb896645.aspx) to trace what file and registry operations will be done during loading of ABQSMABasCoreUtils.dll.
If LoadLibrary is not failed, then you have really an initialization problem of DLLs. Typically the problem exist if a DLL inside of DllMain try use a function from another DLL which is not yet initialized (not yet returns from DllMain). Before one start diagnostic of this problem, we should try to exclude a more simple problems with LoadLibrary.
The ABQSMABasCoreUtils.dll looks like it's importing 64-bit functions. Is your dll also 64-bit? If not, then that's the problem - you cannot mix DLLs compiled for different architectures in the same process.
We recently attempted to break apart some of our Visual Studio projects into libraries, and everything seemed to compile and build fine in a test project with one of the library projects as a dependency. However, attempting to run the application gave us the following nasty run-time error message:
Run-Time Check Failure #0 - The value of ESP was not properly saved across a function call. This is usually a result of calling a function pointer declared with a different calling convention.
We have never even specified calling conventions (__cdecl etc.) for our functions, leaving all the compiler switches on the default. I checked and the project settings are consistent for calling convention across the library and test projects.
Update: One of our devs changed the "Basic Runtime Checks" project setting from "Both (/RTC1, equiv. to /RTCsu)" to "Default" and the run-time vanished, leaving the program running apparently correctly. I do not trust this at all. Was this a proper solution, or a dangerous hack?
This debug error means that the stack pointer register is not returned to its original value after the function call, i.e. that the number of pushes before the function call were not followed by the equal number of pops after the call.
There are 2 reasons for this that I know (both with dynamically loaded libraries). #1 is what VC++ is describing in the error message, but I don't think this is the most often cause of the error (see #2).
1) Mismatched calling conventions:
The caller and the callee do not have a proper agreement on who is going to do what. For example, if you're calling a DLL function that is _stdcall, but you for some reason have it declared as a _cdecl (default in VC++) in your call. This would happen a lot if you're using different languages in different modules etc.
You would have to inspect the declaration of the offending function, and make sure it is not declared twice, and differently.
2) Mismatched types:
The caller and the callee are not compiled with the same types. For example, a common header defines the types in the API and has recently changed, and one module was recompiled, but the other was not--i.e. some types may have a different size in the caller and in the callee.
In that case, the caller pushes the arguments of one size, but the callee (if you're using _stdcall where the callee cleans the stack) pops the different size. The ESP is not, thus, returned to the correct value.
(Of course, these arguments, and others below them, would seem garbled in the called function, but sometimes you can survive that without a visible crash.)
If you have access to all the code, simply recompile it.
I read this in other forum
I was having the same problem, but I just FIXED it. I was getting the same error from the following code:
HMODULE hPowerFunctions = LoadLibrary("Powrprof.dll");
typedef bool (*tSetSuspendStateSig)(BOOL, BOOL, BOOL);
tSetSuspendState SetSuspendState = (tSuspendStateSig)GetProcAddress(hPowerfunctions, "SetSuspendState");
result = SetSuspendState(false, false, false); <---- This line was where the error popped up.
After some investigation, I changed one of the lines to:
typedef bool (WINAPI*tSetSuspendStateSig)(BOOL, BOOL, BOOL);
which solved the problem. If you take a look in the header file where SetSuspendState is found (powrprof.h, part of the SDK), you will see the function prototype is defined as:
BOOLEAN WINAPI SetSuspendState(BOOLEAN, BOOLEAN, BOOLEAN);
So you guys are having a similar problem. When you are calling a given function from a .dll, its signature is probably off. (In my case it was the missing WINAPI keyword).
Hope that helps any future people! :-)
Cheers.
Silencing the check is not the right solution. You have to figure out what is messed up with your calling conventions.
There are quite a few ways to change the calling convetion of a function without explicitly specifying it. extern "C" will do it, STDMETHODIMP/IFACEMETHODIMP will also do it, other macros might do it as well.
I believe if run your program under WinDBG (http://www.microsoft.com/whdc/devtools/debugging/default.mspx), the runtime should break at the point where you hit that problem. You can look at the call stack and figure out which function has the problem and then look at its definition and the declaration that the caller uses.
I saw this error when the code tried to call a function on an object that was not of the expected type.
So, class hierarchy: Parent with children: Child1 and Child2
Child1* pMyChild = 0;
...
pMyChild = pSomeClass->GetTheObj();// This call actually returned a Child2 object
pMyChild->SomeFunction(); // "...value of ESP..." error occurs here
I was getting similar error for AutoIt APIs which i was calling from VC++ program.
typedef long (*AU3_RunFn)(LPCWSTR, LPCWSTR);
However, when I changed the declaration which includes WINAPI, as suggested earlier in the thread, problem vanished.
Code without any error looks like:
typedef long (WINAPI *AU3_RunFn)(LPCWSTR, LPCWSTR);
AU3_RunFn _AU3_RunFn;
HINSTANCE hInstLibrary = LoadLibrary("AutoItX3.dll");
if (hInstLibrary)
{
_AU3_RunFn = (AU3_RunFn)GetProcAddress(hInstLibrary, "AU3_WinActivate");
if (_AU3_RunFn)
_AU3_RunFn(L"Untitled - Notepad",L"");
FreeLibrary(hInstLibrary);
}
It's worth pointing out that this can also be a Visual Studio bug.
I got this issue on VS2017, Win10 x64. At first it made sense, since I was doing weird things casting this to a derived type and wrapping it in a lambda. However, I reverted the code to a previous commit and still got the error, even though it wasn't there before.
I tried restarting and then rebuilding the project, and then the error went away.
I was getting this error calling a function in a DLL which was compiled with a pre-2005 version of Visual C++ from a newer version of VC (2008).
The function had this signature:
LONG WINAPI myFunc( time_t, SYSTEMTIME*, BOOL* );
The problem was that time_t's size is 32 bits in pre-2005 version, but 64 bits since VS2005 (is defined as _time64_t). The call of the function expects a 32 bit variable but gets a 64 bit variable when called from VC >= 2005. As parameters of functions are passed via the stack when using WINAPI calling convention, this corrupts the stack and generates the above mentioned error message ("Run-Time Check Failure #0 ...").
To fix this, it is possible to
#define _USE_32BIT_TIME_T
before including the header file of the DLL or -- better -- change the signature of the function in the header file depending on the VS version (pre-2005 versions don't know _time32_t!):
#if _MSC_VER >= 1400
LONG WINAPI myFunc( _time32_t, SYSTEMTIME*, BOOL* );
#else
LONG WINAPI myFunc( time_t, SYSTEMTIME*, BOOL* );
#endif
Note that you need to use _time32_t instead of time_t in the calling program, of course.
I was having this exact same error after moving functions to a dll and dynamically loading the dll with LoadLibrary and GetProcAddress. I had declared extern "C" for the function in the dll because of the decoration. So that changed calling convention to __cdecl as well. I was declaring function pointers to be __stdcall in the loading code. Once I changed the function pointer from __stdcall to__cdecl in the loading code the runtime error went away.
Are you creating static libs or DLLs? If DLLs, how are the exports defined; how are the import libraries created?
Are the prototypes for the functions in the libs exactly the same as the function declarations where the functions are defined?
do you have any typedef'd function prototypes (eg int (*fn)(int a, int b) )
if you dom you might be have gotten the prototype wrong.
ESP is an error on the calling of a function (can you tell which one in the debugger?) that has a mismatch in the parameters - ie the stack has restored back to the state it started in when you called the function.
You can also get this if you're loading C++ functions that need to be declared extern C - C uses cdecl, C++ uses stdcall calling convention by default (IIRC). Put some extern C wrappers around the imported function prototypes and you may fix it.
If you can run it in the debugger, you'll see the function immediatey. If not, you can set DrWtsn32 to create a minidump that you can load into windbg to see the callstack at the time of the error (you'll need symbols or a mapfile to see the function names though).
Another case where esp can get messed up is with an inadvertent buffer overflow, usually through mistaken use of pointers to work past the boundary of an array. Say you have some C function that looks like
int a, b[2];
Writing to b[3] will probably change a, and anywhere past that is likely to hose the saved esp on the stack.
You would get this error if the function is invoked with a calling convention other than the one it is compiled to.
Visual Studio uses a default calling convention setting thats decalred in the project's options. Check if this value is the same in the orignal project settings and in the new libraries. An over ambitious dev could have set this to _stdcall/pascal in the original since it reduces the code size compared to the default cdecl. So the base process would be using this setting and the new libraries get the default cdecl which causes the problem
Since you have said that you do not use any special calling conventions this seems to be a good probability.
Also do a diff on the headers to see if the declarations / files that the process sees are the same ones that the libraries are compiled with .
ps : Making the warning go away is BAAAD. the underlying error still persists.
This happened to me when accessing a COM object (Visual Studio 2010). I passed the GUID for another interface A for in my call to QueryInterface, but then I cast the retrieved pointer as interface B. This resulted in making a function call to one with an entirely signature, which accounts for the stack (and ESP) being messed up.
Passing the GUID for interface B fixed the problem.
In my MFC C++ app I am experiencing the same problem as reported in Weird MSC 8.0 error: “The value of ESP was not properly saved across a function call…”. The posting has over 42K views and 16 answers/comments none of which blamed the compiler as the problem. At least in my case I can show that the VS2015 compiler is at fault.
My dev and test setup is the following: I have 3 PCs all of which run Win10 version 10.0.10586. All are compiling with VS2015, but here is the difference. Two of the VS2015s have Update 2 while the other has Update 3 applied. The PC with Update 3 works, but the other two with Update 2 fail with the same error as reported in the posting above. My MFC C++ app code is exactly the same on all three PCs.
Conclusion: at least in my case for my app the compiler version (Update 2) contained a bug that broke my code. My app makes heavy use of std::packaged_task so I expect the problem was in that fairly new compiler code.
ESP is the stack pointer. So according to the compiler, your stack pointer is getting messed up. It is hard to say how (or if) this could be happening without seeing some code.
What is the smallest code segment you can get to reproduce this?
If you're using any callback functions with the Windows API, they must be declared using CALLBACK and/or WINAPI. That will apply appropriate decorations to make the compiler generate code that cleans the stack correctly. For example, on Microsoft's compiler it adds __stdcall.
Windows has always used the __stdcall convention as it leads to (slightly) smaller code, with the cleanup happening in the called function rather than at every call site. It's not compatible with varargs functions, though (because only the caller knows how many arguments they pushed).
Here's a stripped down C++ program that produces that error. Compiled using (Microsoft Visual Studio 2003) produces the above mentioned error.
#include "stdafx.h"
char* blah(char *a){
char p[1];
strcat(p, a);
return (char*)p;
}
int main(){
std::cout << blah("a");
std::cin.get();
}
ERROR:
"Run-Time Check Failure #0 - The value of ESP was not properly saved across a function call. This is usually a result of calling a function declared with one calling convention with a function pointer declared with a different calling convention."
I had this same problem here at work. I was updating some very old code that was calling a FARPROC function pointer. If you don't know, FARPROC's are function pointers with ZERO type safety. It's the C equivalent of a typdef'd function pointer, without the compiler type checking.
So for instance, say you have a function that takes 3 parameters. You point a FARPROC to it, and then call it with 4 parameters instead of 3. The extra parameter pushed extra garbage onto the stack, and when it pops off, ESP is now different than when it started. So I solved it by removing the extra parameter to the invocation of the FARPROC function call.
Not the best answer but I just recompiled my code from scratch (rebuild in VS) and then the problem went away.