|
I think I have found a bug in the CString assignement operator of the ATL/MFC 8.
CStringT& operator=( __in_z_opt PCYSTR pszSrc )
The method calculates the length of the required buffer and allocates the buffer. The calculation does not include the terminating null character. Thereafter it calls MultiByteToWideChar, passing the length as the cchWideChar parameter.
The function MultiByteToWideChar returns 0 as failure indication, but this is ignored by the MFC. As a side effect, MultiByteToWideChar fills the output buffer on some platforms like Win32 and Windows CE 5.0 (SH4).
But the Windows CE 5.0 (x86) implementation of MultiByteToWideChar does not fill the buffer. Although the allocated buffer is too small, the bug is not visible on most platforms but on Windows CE 5.0 (x86). Here you get an empty CString after the assignment.
Example:
CStringW s("ABC");
PCWSTR p = (PCWSTR)s;
assert(0 != p[0]); The assert in the code above fails.
|
|
|
|
|
It wont be the first time different platforms have behaved differently. Whether MS will admit its a bug or not, well, they dont like to.
But anyway, contact them and see what they say. There might be a patch already.
==============================
Nothing to say.
|
|
|
|
|
Here is an polar example of the operator's usage  :
{
LPTSTR lpszBuffer(_T("test"));
CString cszDestination;
cszDestination = lpszBuffer; ASSERT(cszDestination[0] == _T('t'));
}
They sought it with thimbles, they sought it with care;
They pursued it with forks and hope;
They threatened its life with a railway-share;
They charmed it with smiles and soap.
|
|
|
|
|
Yes, this calls the assignment constructor directly. It shows the same behavior since it executes the same code.
My example came from a real project.
As long Microsoft is reluctant to fix this I will have to get the debug source code and build my own MFC version. Unfortunately this requires to install this version on all PCs that just wants to assign an (ANSI) string to a string class object. 
|
|
|
|
|
class Filter;
typedef void (Filter::*PFN)();
class Filter
{
public:
void ProcessPart();
void MyProcessPart();
struct Pair_t
{
Pair_t(int d, PFN n) : iid(d), pfn(n) {}
int iid;
PFN pfn;
private:
Pair_t& operator=(const Pair_t&);
};
};
void Filter::ProcessPart()
{
PFN pfn2;
pfn2 = dynamic_cast<Filter*> (&Filter::ProcessPart);
Pair_t(9, pfn2);
}
int _tmain(int argc, _TCHAR* argv[])
{
return 0;
}
I am not able to solve this.
Any help?
Thanks
|
|
|
|
|
You are trying to cast a function address to a pointer to a class, which is illegal.
|
|
|
|
|
Ya I got it.
Thanks Richard MacCutchan
|
|
|
|
|
Function pointer casting has a really crappy syntax in C. I posted on it way back, but so ugly is I have totally forgotten how it goes.
==============================
Nothing to say.
|
|
|
|
|
You've already had help from the clever people here, but it would have helped them if you'd put the error in the posting too.
If you make it easy to help you, more people will try.
Iain.
I am one of "those foreigners coming over here and stealing our jobs". Yay me!
|
|
|
|
|
Iain Clarke, Warrior Programmer wrote: if you'd put the error in the posting too.
Ahem, it's in the original question, as a C++ comment.
|
|
|
|
|
OK, I'll partially let him off then!
But pasting the whole error message is still wise. Do *you* remember E1234 off by heart?
Iain.
I am one of "those foreigners coming over here and stealing our jobs". Yay me!
|
|
|
|
|
Iain Clarke, Warrior Programmer wrote: But pasting the whole error message is still wise.
I agree entirely.
Iain Clarke, Warrior Programmer wrote: Do *you* remember E1234 off by heart?
No, but I can spell MSDN and even Google. 
|
|
|
|
|
Hi
I've been struggling for some time to create a sample app in C++ that can upload and download to/from Google Docs using their 'Resumable Upload API'.
I've got quite a long way using Google's own documentation to implement a REST solution but there is so little example code that when I hit a stumbling block there is nowhere to turn.
Firstly, do you have specific experience with Google Docs?
Secondly, would you be willing to offer some specific feedback on my code?
|
|
|
|
|
Usually things like Google apps (apps or libraries) tend to have forums for their users to interact and ask questions, have you looked to see if such forum exists? You may have better luck with people who have used that specific API there.
|
|
|
|
|
Thx Albert.
I have already tried the Google forums but I'll give them another try.
|
|
|
|
|
Hello,
I have a multi threaded MFC exe built with all its basic classes including CSocket, with TCP connection through mainframe etc. , I want to split it to two different components. ATL service and MFC GUI. Can I do it and can I use the MFC classes in the ATL service.
Thanks in advance,
Eyal
|
|
|
|
|
hello guys... Is it possible to encrypt voice just like we can encrypt data. Encryption methodology can be any like: reversing the voice. For example, If I speak hello then I want it to be converted to olleh. Thnx
This world is going to explode due to international politics, SOON.
|
|
|
|
|
You can encrypt any bit stream using the various libraries available for the purpose. To do what you are asking would require analysis of the stream and separation at each character of each word, which is impossible in spoken sounds.
Try saying "hello" with different stresses and see if you can determine the exact point where each letter starts and ends; do you hear one or two 'l's?
|
|
|
|
|
As long as the data is digitized (using PCM or some other encoding technique) it can be encrypted. Speech recognition is a completely different issue, and quite frankly I would advice you to not try to do that at this point (that's a pretty involved task). I would start with:
0. Learn how to turn voice into digital data using PCM (learn capture and playback techniques).
1. Learn how to use encryption (using a library is the fastest method).
modified 12-Jun-12 11:17am.
|
|
|
|
|
I have a property page with at least 2 edit controls.
When the user presses enter on the first edit box, I want the 2nd edit box to become active and gain focus. I have tried:
void PropPageHello::OnKeyDown(UINT nChar, UINT nRepCnt, UINT nFlags)
{
CPropertyPage::OnKeyDown(nChar, nRepCnt, nFlags);
}
If you put a break pointin this function, it is never hit. I am not sure how to capture the enter key and set the focus of the next item.
modified 11-Jun-12 14:59pm.
|
|
|
|
|
instead of OnKeyDown, try PreTranslateMessage:
BOOL CMyDlg::PreTranslateMessage(MSG* pMsg)
{
if(pMsg->message==WM_KEYDOWN)
{
if(pMsg->wParam==VK_RETURN)
{
}
}
return CDialog::PreTranslateMessage(pMsg);
}
|
|
|
|
|
|
Howdy'
(question at the end)
I'm looking into refactoring our code to be able to use all of the new MFC (VS2008) expansion UI goodies (custom toolbars, application framework, ...)
One of those goodies is the customize dialog (to be able to create new toolbars, move menu and toolbar buttons around).
From what I understand, the dialog uses resource ID in the 16000-17000(-ish) (afxribbonres.h )
Those ID come into conflict with some of our ID (mostly bad string in the new UI), no biggies, I will fix/renumber our ID.
Question:
Is there a proper way to handle those kind of issues ? a way to tell me at compilation time that I have resource ID conflicts when including external resource header file ?
Thanks.
Max.
Watched code never compiles.
|
|
|
|
|
Maximilien wrote: to be able to use all of the new MFC (VS2008) expansion UI goodies (custom toolbars, application framework, ...)
New?
I am not sure if the compiler can tell you about resource ID conflicts, but if you check this note[^] it gives some good guidance.
|
|
|
|
|
Richard MacCutchan wrote: New?
New as in "MFC Feature Pack 2008"  (4 years!)
The compiler does not report ID conflicts, it all builds nicely.
Anyway, I will clean up my hundreds of resources in the offending range.
Watched code never compiles.
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
