|
I am learning Program COM by ATL,I build a inpro-server dll and add a com object ComputerInfo,I only add a method to IComputerInfo and not to any thing just as bellow:
interface IComputerInfo : IDispatch
{
[id(1), helpstring("method Test")] HRESULT GetCPU(char* bufName,const int* size);
};
STDMETHODIMP CComputerInfo::GetCPU(char* bufName,const int* size)
{
// TODO: Add your implementation code here
return S_OK;
}
Compiler shows:
ATLFirst.cpp
error C2259: 'CComObject<class ccomputerinfo="">' : cannot instantiate abstract class due to following members:
atlcom.h(1823) : while compiling class-template member function 'long __stdcall ATL::CComCreator<class atl::ccomobject<class="" ccomputerinfo=""> >::CreateInstance(void *,const struct _GUID &,void ** )'
f:\microsoft visual studio\vc98\atl\include\atlcom.h(1827) : warning C4259: 'long __stdcall IComputerInfo::GetCPU(unsigned char *,const int *)' : pure virtual function was not defined
if I have change the metho to GetCPU(unsigned char* bufName,const int* size)
all right,it works well,why???
|
|
|
|
|
I think this is a problem (bug?) in the IDL compiler, that generated the unsigned char* signature instead the char*
Anyway, if you want to be able to call your component from e.g. Visual Basic or something, it's better to use BSTR type for strings. This is basically counted unicode string (Basic STRing). Then you can cope with that using helper classes CComBSTR from ATL or _bstr_t from comdef.h
Hope this helps
|
|
|
|
|
thanks a lot.and I also have a question.
unsigned char buf[]="hello";//OK
strcpy(buf,"hello");//Error:because buf is not char*
and if I forced buf to char* and use MessageBox to show it,I found many rubbish code.I don't know the advantage of the unsigned char*,could you tell me?
|
|
|
|
|
unsigned char is in equal to BYTE type from windows header. Difference between these is stated in the name of the type - both are 1 byte of size, but char is signed and unsigned char is... unsigned (unless you switch the C++ compiler to treat char as unsigned).
because in C++ the type checking is quite strict, you cannot implicitly convert unsigned to signed and vice versa. You have to state explicitly, that you really want to do it. (e.g. by casting)
Just an advice - don't use the strcpy and similar functions. Find and stick to some string wrapper - be it std::string from stl libraries or CString from MFC/ATL. This would save you lot of troubles when looking for bugs inside string manipulation (forgotten ending \0, buffer overruns or whatever)
|
|
|
|
|
Hi all,
I'm facing a big problem with strings in ATL component.
it need to append to same string in a function.
that CComBSTR object is passed to a function by reference which appends some data to it.
but in appending steps at certain stages it show increase in memory usages.
Is this memory leak.
how to work around this problem.
thanks,
Prasad
|
|
|
|
|
Can you specify how you determined memory usage has increased ?
Also check whether this link has something to say.INFO :q139071
There is no spoon.
mail
|
|
|
|
|
in my component there is memory leak.
when i was debugging application, i found in task manager that append statement of CComBSTR causes some increase in memory usage
|
|
|
|
|
I'm sure someone else will point this out...
An increase in memory usage is not necessarily an indication that the program leaks memory. CComBSTR may use temporary heap space while it does the append operation. This may cause a change in the memory size of your app. When the memory is freed (via delete or free) this is not necessarily reflected immediately by the memory being marked as free from the OS point of view.
If you do repeated appends (say, 1000 in a loop), and it still increases, that may be a better indication.
Steve S
Developer for hire
|
|
|
|
|
Hi everyone.
Recently I've began to study WTL programming and I 've got a question. Please explain me how can I create a colour bitmap from array of bits? Not colour bitmap I create like this:
HBITMAP btmp;
LPBITMAPINFO pbi;
LPBITMAPINFOHEADER pbmph;
bool CDisplay::initBmp(byte* Img)
{
if(pbi != NULL)
delete [] pbi;
if(btmp != NULL)
DeleteObject(btmp);
int headerSize = sizeof(BITMAPINFOHEADER) + sizeof(RGBQUAD) * 256;
try
{
pbi = (LPBITMAPINFO)new char [headerSize];
pbmph = &(pbi->bmiHeader);
memset(pbmph, 0, headerSize);
pbmph->biSize = sizeof(BITMAPINFOHEADER);
pbmph->biPlanes = 1;
pbmph->biCompression = BI_RGB;
pbmph->biClrUsed = 256;
pbmph->biClrImportant = 256;
pbmph->biBitCount = 8;
// fill RGBQUAD
LPRGBQUAD pRGBQ = (LPRGBQUAD)((LPSTR)pbmph + pbmph->biSize);
for( int i = 0; i < 256; i++ )
{
pRGBQ[i].rgbBlue =
pRGBQ[i].rgbGreen =
pRGBQ[i].rgbRed = i;
}
}
catch(...)
{
return false;
}
pbmph->biWidth = MaxX[0];
pbmph->biHeight = MaxY[0];
pbmph->biSizeImage = MaxX[0]*MaxY[0];
btmp = CreateDIBitmap(GetDC(), pbmph, CBM_INIT, Img, pbi, DIB_RGB_COLORS);
return true;
}
How create colour bitmap?
P.S.: Sorry for my english if I wrote with errors.
Nulla dies sine linea !!!
|
|
|
|
|
You might want to look at CreateDIBSection. This is similar to what you already have (it will return a HBITMAP handle and a pointer to the bits). You can directly modify the bits via that pointer and use the handle for selecting the object into DCs.
If you simply want to create a 24-bit colour image, you don't need to create palette information in the BITMAPINFOHEADER (set those values to 0). Then simply write an RGB TRIPLE (Blue, Green, Red) for each pixel using the pointer. Remember you will need to pad each scanline to the DWORD boundary. If you want to write a 256-colour (paletted) image, you will need to provide a colour table. Check out MSDN documentation for that information.
Another gotcha is that if you want the bitmap to be "top-down" i.e. pixel (0,0) is in the top-left corner, you must specify a NEGATIVE height value in the BITMAPINFOHEADER structure.
Hope that helps
|
|
|
|
|
TNX, That helped me.
I did it.
Nulla dies sine linea !!!
|
|
|
|
|
Hi,
I am trying to use a COM component in MFC client app. I call one Method of a this component but it throws and error with HRESULT as 0x80020008. The error code I verified corresponds to FACILITY_DISPATCH and DISP_E_BADVARTYPE.
The problem is - it works on another machine but not on my machine. Also, I am using TLH, TLI file generated for this component using #import directive.
Please provide me with some solution to this.
Thanks in advance.
Have a great day ahead!
Regards,
Sohail Kadiwala
(My Blog - http://blogs.wdevs.com/sohail/[^])
modified 21-Apr-21 21:01pm.
|
|
|
|
|
Difficult for saying without further info. But from the description I think, that you're passing some VARIANT parameter into the function, or the function expects it. (or some other VARIANT conversion takes place on the way).
Then for the problem one-machine-works-other-not I would check the values passed to the function (or possibly it can be also some operation inside the COM object) for some regional setting specific parts - number with decimal dot, dates... If you pass e.g. a number 1.2 as a VARIANT type string (VT_BSTR) on some machines it can be a working construct, because they know that decimal dot is a separator and then can safely convert it to required number. On other machine, where is comma set as separator, the conversion can fail.
But these are just thoughts, without more details it's difficult to tell what can be wrong.
|
|
|
|
|
Hi geo
Thanks for information. Look I am not passing anything myself. The Method got absoluetly no [IN] parameter.
Also I said in my post that I am using TLH and TLI files and these are generated using #import "component.dll"
Even I have only limited information.
If possible try giving me some info regarding this error. Also, I don't have source of the Component
Have a great day ahead!
Regards,
Sohail Kadiwala
(My Blog - http://blogs.wdevs.com/sohail/[^])
modified 21-Apr-21 21:01pm.
|
|
|
|
|
Hmm, that would be quite difficult to debug - I would check, if there is difference between these two machines in regional settings (for user as well as for the system). Also take into account what the method call have to do (can help to focus a bit - if it should work with date/time or numbers or whatever).
Also possibility is that it calls internally some other component and on one machine is different version that on the other one.
Anyway, if there's no support neither source code for the component, I would think about dumping it and trying to find/develop somethin what does the same job but more reliable, because now, you can spend some few days with locating the bug and tomorrow at client's site there can be other problems similar or worse than this one.
|
|
|
|
|
Thanks Geo,
I'll take a look at things you have mentioned. Also the ReDesigning stuff 
Have a great day ahead!
Regards,
Sohail Kadiwala
(My Blog - http://blogs.wdevs.com/sohail/[^])
modified 21-Apr-21 21:01pm.
|
|
|
|
|
I found many examples about DirectoryWatch, FileWatch. I need help regarding the moment of closing of the file. Suppose some operation on the file is going on by some application. I want to add some information at the end of the file but immediately after it is closed by the application. So my application must be able to trace the closing of the file.
Moyeen
Sr. Software Design Engineer
|
|
|
|
|
look @ ReadDirectoryChangesW with FILE_NOTIFY_CHANGE_LAST_WRITE flag set. This is the nearest option I can imagine to help you quite easily.
Problem is that closing the file handle can mean nothing for the underlying filesystem in some situations, and I don't know any API that would allow you go as deep as to file locking mechanism that CAN be influenced by simple file closing (not saying that the behavior of such a functions can be different for NTFS or FAT32 or network drive)
Maybe you can track the file handles directly, but this will need a great experience in low level system programming (although nothing impossible I think) but it can also cause bugs in other applications, because you have to dig in other processes.
|
|
|
|
|
hi all,
As I understood it vectors are bound safe, i.e. you can't go past the end of the vector. However I've found some code which does just that!
the vector is defined as an "vector < vector < int > >" and although when you try to access an out of bound area on the first part you get an error, if you do the same on the second part it allows you. An example is if the first vector has 10 elements (i.e. 10 x vector < int > ) and the second part has 10 elements (i.e. 10 x int). If you do "myVector[i][j] = 6 ;" where i=20 and j=5 then you get a crash, whereas if i=5 and j=20 you don't! 
Can anyone tell me if this is standard behaviour or could it be a compiler 'feature' (I'm using MS VC++ 6.0 with (iirc) SP5).
cheers,
Andy
|
|
|
|
|
You must access the vector elements using the at() member function to get bounds checking. Using the [] operator to access out of bounds elements causes undefined behavior.
|
|
|
|
|
Also keep in mind that at() throws on failure. Good to know in beforehand 
--
Weiter, weiter, ins verderben.
Wir müssen leben bis wir sterben.
I blog too now[^]
|
|
|
|
|
Can anyone tell how to pass an array from an avent in case of ATL COM DLL.Awaiting for respense eagerly.Plz respond as soon as possible.
Akshay Dave
|
|
|
|
|
use SAFEARRAY
see in MSDN CreateSafeArray and related functions
|
|
|
|
|
Hi Akshay ,
i just adding to real Monkey advice,
just declare the return element VARIANT, and using SAFEARRAY return an array to Client.
-----------------------------
"I Think It Will Help"
-----------------------------
Alok Gupta
visit me at http://www.thisisalok.tk
|
|
|
|
|
Hi Alok,Thanx 4 ur reply.I had tried out 4 what u have suggested.But I think I'm not able to implement VARAINT SAFEARRAY it is giving warning in Proxy file it generates 4 event.It says"Forcing a tagSafeArray value to Bool true
or false".Plz reply me with some codes of VB and Vc if possible.My email-id is avd9683@rediffmail.com.
Awaiting 4 ur reply soon.
AKshay Dave
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
