C / C++ / MFC

Re: change entry point address

31-Aug-05 10:54

Hi Spirit,

First id like to declare that Ive never used the functions that you're using
in this code snippet but here's my two bobs worth.

1) It looks like MapViewOfFile is returning NULL (which it does in case of error). (so pDosHeader->e_lfanew evaluates to 0x3c... sounds feasible?)

2) Reading to doco it looks like the "dwDesiredAccess" flag isnt "addable"
(Ths do saye "This parameter can be one of the following values").

It also says that FILE_MAP_WRITE by iself gives RW access.

Try playing with the MapViewOfFile args and check the returned value for NULL maybe

Cheers

Blake Miller31-Aug-05 10:59

Blake Miller

31-Aug-05 10:59

Maybe because in MapViewOfFile you left the last parameter at 0 (dwNumberOfBytesToMap) instead of setting it to size of file?

Dialog List Problem

bugDanny31-Aug-05 7:54

31-Aug-05 7:54

I’m trying to use a list box control in a windows app I’m writing. I put on the control on my dialog resource, I assign it a member variable: as a control, CListBox. But in my code, when I try to use the function AddString (i.e. dlg.m_variableName.AddString(“Some Letters”) ) I get a debug assertion failure. Maybe I’m missing a step here? Is there more initialization I have to do?

I don’t really need the selectability of a list box. I tried just making it an edit box with a scroll bar, but I can’t get each item on a new line.

Danny

C^hris L^osinger31-Aug-05 9:11

C^hris L^osinger

31-Aug-05 9:11

where does the assert happen ?

Cleek | Image Toolkits | Thumbnail maker

bugDanny31-Aug-05 9:16

31-Aug-05 9:16

Chris Losinger wrote:
where does the assert happen ?

The simple answer? Somewhere in the AddString method.
The more complicated answer? The debug pop up tells me the assertion failed at afxwin2.inl at line 669.

Danny

C^hris L^osinger31-Aug-05 9:24

C^hris L^osinger

31-Aug-05 9:24

_AFXWIN_INLINE int CListBox::AddString(LPCTSTR lpszItem)
	{ ASSERT(::IsWindow(m_hWnd)); return (int)::SendMessage(m_hWnd, LB_ADDSTRING, 0, (LPARAM)lpszItem); }

it's failing because the CListBox isn't a window (yet). i'm guessing you're calling AddString before the dialog has had a chance to create the list box and/or associate it with your member variable.

put a breakpoint on your AddString and one in your dialog's OnInitDialog. if the AddString breakpoint hits first, you need to move it to after the dialog has been initialized.

Cleek | Image Toolkits | Thumbnail maker

bugDanny31-Aug-05 9:31

31-Aug-05 9:31

Thanks!

Danny

bugDanny31-Aug-05 9:18

31-Aug-05 9:18

Alternatively, the release version works, or rather, survives. It works with the exception of the text that I wanted to put in the list box is not there.

Danny

Gary R. Wheeler3-Sep-05 6:13

Gary R. Wheeler

3-Sep-05 6:13

The debug version ASSERTs because the window hasn't been created yet. In the release version, the ASSERT doesn't fire, and the SendMessage call silently fails. In this case, the control eventually gets created, but the string you're looking for isn't there due to the earlier SendMessage failure.

Software Zen: delete this;

David Crow31-Aug-05 16:10

David Crow

31-Aug-05 16:10

bugDanny wrote:
But in my code, when I try to use the function AddString

Make sure you are doing this in the dialog's OnInitDialog() method, after the default method has been called.

"One must learn from the bite of the fire to leave it alone." - Native American Proverb

File Size Limits

carks31-Aug-05 6:20

31-Aug-05 6:20

Im trying to read a large text file with a CStdioFile object (the file's around 2.3 Gb, OS Windows XP, File System NTFS).

After reading for a while an exception is thrown during a CStdioFile::ReadString call which says something like "Controler Invalid"

I simplified the code so that its doing nothing more than loopìng around the ReadString call, and the error persists.

Could there be an inherent limit on the size of a file readable using CStdioFile (I dont belive that 2.3 Gb is too big for NTFS ?!?)

Thanks,

Dave

JWood31-Aug-05 6:41

JWood

31-Aug-05 6:41

You may be running into some unexpected 32bit dword limitations, not necessarily a file os problem, but a file pointer limitation.

carks31-Aug-05 6:54

31-Aug-05 6:54

That sounds something like whats happening, but does anybody know of any settable limmitations of this type that exist?

PJ Arends31-Aug-05 7:05

PJ Arends

31-Aug-05 7:05

Are you using VC6? if so then the file pointer used is a 32 bit value, with a limit of 2GB. AFAIK if you are using VC7+ then the file pointer is a 64 bit value.

If you want to get around this limit in VC6 then do not use the MFC classes (which wrap the fopen, fread, etc CRT functions) and use the SDK functions CreateFile, ReadFile etc.

"You're obviously a superstar." - Christian Graus about me - 12 Feb '03

"Obviously ??? You're definitely a superstar!!!" - mYkel - 21 Jun '04

"There's not enough blatant self-congratulatory backslapping in the world today..." - HumblePie - 21 Jun '05

Within you lies the power for good - Use it!

carks31-Aug-05 7:21

31-Aug-05 7:21

Yes thats it PJ!

I am using VC6.

(I guess its time to shell out for an upgrade!)

Thanks so much for the help Smile | :)

Bob Stanneveld31-Aug-05 7:54

Bob Stanneveld

31-Aug-05 7:54

PJ Arends wrote:
do not use the MFC classes (which wrap the fopen, fread, etc CRT functions)

Seriously? I always though that MFC didn't wrap CRT functions, but rather the WIN API directly! The CRT file API's (I don't know about the others) are just a layer on top of the OS's API's, which are CreateFile, ReadFile, etc..

I don't know about the MFC shipped with VC6, but the newer MFC version 8, does wrap the window's API's...

Behind every great black man...
... is the police. - Conspiracy brother

Blog[^]

Blake Miller31-Aug-05 8:37

Blake Miller

31-Aug-05 8:37

From VC 6.0 MFC Source (FILECORE.CPP Line 111)
Yes it wraps Win32 API, not standard C runtime calls...

BOOL CFile::Open(LPCTSTR lpszFileName, UINT nOpenFlags,<br />
	CFileException* pException)<br />
{<br />
<br />
...<br />
<br />
	// attempt file creation<br />
	HANDLE hFile = ::CreateFile(lpszFileName, dwAccess, dwShareMode, &sa,<br />
		dwCreateFlag, FILE_ATTRIBUTE_NORMAL, NULL);<br />
	if (hFile == INVALID_HANDLE_VALUE)<br />
	{<br />
		if (pException != NULL)<br />
		{<br />
			pException->m_lOsError = ::GetLastError();<br />
			pException->m_cause =<br />
				CFileException::OsErrorToException(pException->m_lOsError);<br />
<br />
			// use passed file name (not expanded vesion) when reporting<br />
			// an error while opening<br />
<br />
			pException->m_strFileName = lpszFileName;<br />
		}<br />
		return FALSE;<br />
	}<br />
	m_hFile = (HFILE)hFile;<br />
	m_bCloseOnDelete = TRUE;

Bob Stanneveld31-Aug-05 8:47

Bob Stanneveld

31-Aug-05 8:47

Just as I expected. It would be just silly for MS to use some portable API that is a layer on top of their own API, instead of their own API.

Thanks for the info! Big Grin | :-D

Behind every great black man...
... is the police. - Conspiracy brother

Blog[^]

PJ Arends31-Aug-05 10:10

PJ Arends

31-Aug-05 10:10

I guess I should read the MFC source instead of MSDN[^]

[quote]
A CStdioFile object represents a C run-time stream file as opened by the run-time function fopen.
[/quote]

Bob Stanneveld31-Aug-05 11:00

Bob Stanneveld

31-Aug-05 11:00

So much for the accoutibility of MSDN...

Behind every great black man...
... is the police. - Conspiracy brother

Blog[^]

Blake Miller31-Aug-05 13:00

Blake Miller

31-Aug-05 13:00

Well, I DID look at the CFile code. Who knows, maybe the CStdioFile object does something different? Not likely, but this IS MFC ... OMG | :OMG:

Toby Opferman31-Aug-05 14:35

Toby Opferman

31-Aug-05 14:35

Generally what I think that this is referring to is that this object *IS* the C run time and behaves the same. In the open it does call CFile::Open:

0:000> u MFC42!CStdioFile::Open
MFC42!CStdioFile::Open:
73e2c96c 8bff             mov     edi,edi
73e2c96e 55               push    ebp
73e2c96f 8bec             mov     ebp,esp
73e2c971 53               push    ebx
73e2c972 8b5d0c           mov     ebx,[ebp+0xc]
73e2c975 56               push    esi
73e2c976 57               push    edi
73e2c977 ff7510           push    dword ptr [ebp+0x10]
0:000> u
MFC42!CStdioFile::Open+0xe:
73e2c97a 8bc3             mov     eax,ebx
73e2c97c 25ffbfffff       and     eax,0xffffbfff
73e2c981 50               push    eax
73e2c982 ff7508           push    dword ptr [ebp+0x8]
73e2c985 8bf9             mov     edi,ecx
73e2c987 33f6             xor     esi,esi
73e2c989 897710           mov     [edi+0x10],esi
73e2c98c e8e043fbff       call    MFC42!CFile::Open (73de0d71)

And that calls CreateFile, however right afterwards they do create a standard handle:

73e2c9db 75d2             jnz     MFC42!CStdioFile::Open+0x43 (73e2c9af)
73e2c9dd ebd7             jmp     MFC42!CStdioFile::Open+0x4a (73e2c9b6)
73e2c9df c644050c74       mov     byte ptr [ebp+eax+0xc],0x74
73e2c9e4 40               inc     eax
73e2c9e5 51               push    ecx
0:000>
MFC42!CStdioFile::Open+0x7a:
73e2c9e6 ff7704           push    dword ptr [edi+0x4]
73e2c9e9 c644050c00       mov     byte ptr [ebp+eax+0xc],0x0
73e2c9ee ff150467e773 call dword ptr [MFC42!_imp___open_osfhandle (73e76704)]
73e2c9f4 83f8ff           cmp     eax,0xffffffff
73e2c9f7 59               pop     ecx
73e2c9f8 59               pop     ecx

Look, most other functions are wrappers around C runtime, as an example, look here:

0:000>  u  MFC42!CStdioFile::WriteString
MFC42!CStdioFile::WriteString:
73e2cb01 8bff             mov     edi,edi
73e2cb03 56               push    esi
73e2cb04 8bf1             mov     esi,ecx
73e2cb06 ff7610           push    dword ptr [esi+0x10]  <-- FILE *
73e2cb09 ff74240c         push    dword ptr [esp+0xc]
73e2cb0d ff151867e773     call    dword ptr [MFC42!_imp__fputs (73e76718)]
73e2cb13 83f8ff           cmp     eax,0xffffffff

In WriteString, fputs is called and it gets the FILE * from the "this" pointer at offset 16. So, CStdioFile calls some standard APIs, some C runtime APIs and it in itself essentially re-implements some of the functionality that would have been done had you really called fopen() in order to create a compatible FILE *.

Read calls fread:

MFC42!CStdioFile::Read:
73e2ca4b 837c240800       cmp     dword ptr [esp+0x8],0x0
73e2ca50 56               push    esi
73e2ca51 8bf1             mov     esi,ecx
73e2ca53 7504             jnz     MFC42!CStdioFile::Read+0xe (73e2ca59)
73e2ca55 33c0             xor     eax,eax
73e2ca57 eb62             jmp     MFC42!CStdioFile::Read+0x70 (73e2cabb)
73e2ca59 55               push    ebp
73e2ca5a 57               push    edi
0:000> u
MFC42!CStdioFile::Read+0x10:
73e2ca5b ff7610           push    dword ptr [esi+0x10]  <-- Again, FILE *
73e2ca5e ff742418         push    dword ptr [esp+0x18]
73e2ca62 6a01             push    0x1
73e2ca64 ff74241c         push    dword ptr [esp+0x1c]
73e2ca68 ff151067e773     call    dword ptr [MFC42!_imp__fread (73e76710)]

Notice:

CStdioFile::m_pStream
Remarks

The m_pStream data member is the pointer to an open file as returned by the C run-time function fopen. It is NULL if the file has never been opened or has been closed.

In the documentation. While "fopen" may not have been called, they basically implemented fopen themsevles on top of CFile::Open (that calls CreateFile). Some of their APIs then call standard C Runtime and this public data member (most likely the same one I showed in the dissassembly) is exported as a C runtime compatible pointer. So, whether they use the C Runtime internally or not is not the point, the object itself exports the appropriate data structures and behaves idential to the c runtime implementation.

8bc7c0ec02c0e404c0cc0680f7018827ebee