|
Hi,
I have an SDI app that puts stuff on the view with an OnDraw() override. It updates automatically whenever there is any Document activity but I would like to trigger a screen refresh after certain non-document events, like changes in other dialogs. I don't know how to piggyback another message map on top of OnDraw without screwing it up. Can you offer any clues?
I thought this code would work but no go: "illegal call of non-static member function". It's a virtual override generated by the ClassWizard.
CDC *pDC;
pDC = pMain->GetDC();
CAppView::OnDraw( pDC );
Thanks in advance for any advice.
|
|
|
|
|
You have to have an instance of a class to call functions of it (unless it's staic, as the error message notes).
OnDraw should not be called directly even if you have a pointer to the view, however. You should invalidate it and redraw, there are several mechanisms for this. Personally, I prefer RedrawWindow.
You can either pass a pointer to the view to the different unrelated classes, which might not be a good idea structurewise, or redraw the complete main window. This can be accomplished by calling AfxGetMainWnd()->RedrawWindow();
|
|
|
|
|
Thank you very much for your fast and helpful repsonse. RedrawWindow works nicely.
|
|
|
|
|
I want to create dial-in connection as dial-up connection,
I can create dial-up connection by using RasSetEntryProperties
Can you tell me which funciton be used for dial-in creation.
Thanks a lot!
|
|
|
|
|
char [10] = 10 bytes
unsigned int= 4 bytes
char [1] = 1 byte
MAJOR = 4 bytes
10 + 4 + 1 + 4 = 19 bytes
but it outputs 24 bytes!!
typedef enum {
CS, MIS, ME, PHI
}MAJOR;
typedef struct STUDENT_t {
char name[10];
unsigned int ID;
char sex[1];
MAJOR major;
}STUDENT;
int main(void)
{
printf("%d\n", sizeof(STUDENT));
return 0;
}
-----------------------------
C++ without virtual functions is not OO. Programming with classes but without dynamic binding is called "object based", but not "object oriented".
|
|
|
|
|
I'm just stabbing this one, but the other 5 bytes are for the struct, sepearately its 19bytes but within a struct, you allow a little more room.
Discovery consist of seeing what everybody has seen and thinking what nobody has thought -- Albert Szent-Györgyi
Name the greatest of all the inventors: accident --Mark Twain
|
|
|
|
|
I think you have the Struct Member Alignment set to 4 bytes (see option under Project/Settings/c++/code generation). For the struct you have shown the total size of the
struct would then be 24 bytes.
As it says in MSDN:
The Struct Member Alignment (/Zpn) option controls how the members of a structure are packed into memory and specifies the same packing for all structures in a module. When you specify this option, each structure member after the first is stored on either the size of the member type or n-byte boundaries (where n is 1, 2, 4, 8, or 16), whichever is smaller.
|
|
|
|
|
In a word: boundary alignment. ID is an unsigned int and MAJOR, an enum, also defaults to an int datatype. Ints are 4-byte fields that are naturally aligned on four byte boundarys. Therefore the compiler places the int fields on four byte boundarys. Doubles go on an eight byte boundary.
If the struct does not have the fields ordered in descend size of datatypes, (doubles, ints, short ints and finally char/char array), then the compiler silently inserts unreachable filler alignment bytes. Doubles, ints, etc. are boundary aligned to improve register load performance. On certain types of machines fetching unaligned data can cause an execption.
You can use a #pragma to modify the default boundary alighment.
You can see the result of what the compiler has done by generating a listing that details the struct.
Offset length Field
0 10 name
10 2 unreachable filler inserted by the compiler
12 4 ID
16 1 sex
17 3 unreachable filler inserted by the compiler
20 4 major
Total length=24
Sam
|
|
|
|
|
hi Alex,
Visual C++ ( for that matter most C++ compilers ) has got some thing
called structure alignment. For the efficient access of the memory they align varibles at the DWORD boundary.
Even though name is taking only 10 bytes. The value of ID is from
byte 12 to 16 . Even though sex takes only 1 byte , Major starts only at the 20th byte. And the size will be rounded to nearest DWORD (that is 24 )
u can use a pragma to make it one byte aligned.
#pragma pack(1)
typedef struct STUDENT_t {
char name[10];
unsigned int ID;
char sex[1];
MAJOR major;
}STUDENT;
#pragma pack(4)
now the size will be 19 as u said.
Praseed Pai
www.praseedpai.com
|
|
|
|
|
Thanks guys for all your replies.
Hi Praseed,
That's interesting. Would you tell me what does the following do?
Also, why would the nearest DWORD be 24 bytes?
#pragma pack(1)
...
#pragma pack(4)
Another question is on the DWORD. Isn't DWORD a macro for an
integer type? When do we use DWORD?
Thanks
-----------------------------
C++ without virtual functions is not OO. Programming with classes but without dynamic binding is called "object based", but not "object oriented".
|
|
|
|
|
hi Alex,
DWORD is not a macro for integer type.
i feel it is a typedef
typedef unsigned long DWORD.
Win32 code is running on a 32-bit processor . So accessing memory at
DWORD ( 4 byte aligment ) will speed things up. Visual C++ is padding to the nearest DWORD boundary to maximize performance when u have declared
something like this.
STUDENT x[10];
#pragma pack(1) makes the structure aligment into 1 and it will behave the
way u expected it to be .
ideal usage should be
#pragma pack(push) //push the current settings into pragma stack
#pragam pack(1)
// Declare the student structure here
#pragma pack(pop)
Hope this helps
Praseed Pai
www.praseedpai.com
|
|
|
|
|
Hi praseed,
Thanks for your reply.
So, when do we use DWORD in C/C++? There is a lot of
DWORD value in the registry in the windows OS, but not
sure when to use it in programming.
Thanks
-----------------------------
C++ without virtual functions is not OO. Programming with classes but without dynamic binding is called "object based", but not "object oriented".
|
|
|
|
|
Hi ,
This is the defenition of DWORD in one of the microsoft's header file.
typedef unsigned long DWORD;
usigned long is only 4 bytes (32 bits ) on Intel and AMD 32 bit platform.
( ur confusion might have started from the fact that java long is 8 bytes )
if u use DWORD (win32 data types ) , ur code can be ported to 64-bit platforms without much problem ( at least that is what ms says ).
i feel , microsoft will make sure that DWORD is always 32 bits . Where as they cannot say the thing sure about c types like int , long etc. (with IA-64 or AMD-64 !)
praseed pai
www.praseedpai.com
|
|
|
|
|
By default, data items are aligned to a boundary that matches their size - not necessarily a DWORD. A short will be aligned on a two-byte boundary, a char on a one-byte boundary, and a double on an eight-boundary.
Also, the best way to use the pack pragma is to save and restore the previous alignment. This is done as follows :
#pragma pack(push, 1) // saves old packing and sets it to 1
...
#pragma pack(pop) // restores previous packing
__________________________________________
a two cent stamp short of going postal.
|
|
|
|
|
Hello,
I am having a problem with the CreateProcess method:
The following works, starting a Notepad
PROCESS_INFORMATION pi;
STARTUPINFO si;
ZeroMemory(&si, sizeof(STARTUPINFO));
si.cb = sizeof(si);
si.wShowWindow = SW_SHOW;
ZeroMemory(&pi, sizeof(pi));
CreateProcess(NULL, "Notepad.exe", NULL, NULL, FALSE,
0, NULL, NULL, &si, &pi);
but fails, if I replace the Notepad.exe with "iexplore.exe" for IE and "wordpad.exe" for WordPad. Any reason for this difference?
Best regards,
Paul.
Jesus Christ is LOVE! Please tell somebody.
|
|
|
|
|
One possible reason is that notepad.exe is in the windows directory.
Kuphryn
|
|
|
|
|
That may very well be the problem. The MSDN docs says "The function will not use the search path.".
--
Ich bin Joachim von Hassel, und ich bin Pilot der Bundeswehr.
Welle: Erdball - F104-G Starfighter
|
|
|
|
|
Nice information, I will look into it.
Thanks for the support.
Best regards,
Paul.
Jesus Christ is LOVE! Please tell somebody.
|
|
|
|
|
What does GetLastError() (or @ERR in the debug watch window) say?
--
Ich bin Joachim von Hassel, und ich bin Pilot der Bundeswehr.
Welle: Erdball - F104-G Starfighter
|
|
|
|
|
Not tried it, will do it this evening.
Thanks for the support.
Best regards,
Paul.
Jesus Christ is LOVE! Please tell somebody.
|
|
|
|
|
Hello,
Does anyone know if there is a comparable function in the Windows environment to the 'dig' function from Bind in the Unix/Linux environment?
I did get the dig.exe binary from isc.org, and the support libraries (libdns.dll libeay32.dll libisc.dll), but this seems convoluted to have to include these with my application.
I guess I could use nslookup, but I don't know how long it will be supported since it has been deprecated.
Thanks,
Murrah Boswell
|
|
|
|
|
My plan is actually quite simple. . . I want to create a 4-pane symmetrical splitter in the mainframe client area between a docked toolbar and a status bar for an SDI application. In order to properly set the initial pane sizes, I must use GetWindowRect for the status and tool bars to find the available client rectangle. Everything I have read thus far indicates that the splitter should be created in the overridden OnCreateClient funtion called by the framework. Now for my problem. . .
I cannot access these functions for the controls in the OnCreateClient function. I get assertion errors when I try to get their dimensions. I also cannot get the correct dimension in the OnCreate function. For testing purposes, I simply copied my code for getting these rectangles into the OnPaint function and accessing the correct sizes is no problem. Therefore, I can only assume that I cannot access them in the other functions because the creation of the controls is not complete until AFTER the OnCreate (in which the controls are created) and OnCreateClient functions have returned.
What is the appropriate method for getting the control dimensions with which to initialize the correct pane size in my splitter window? I noticed that in one article the calls to GetWindowRect for the controls were placed in the OnSize function. But, when OnSize is called, I want to set the panes to occupy the same proportion of the client area space as they did before the resize. I must, at some point, initialize their size to 1/4 the usable client area, but where it would be best to do this. . .I have no idea, but I know that I can't do it in OnCreateClient where it makes the most sense.
|
|
|
|
|
You should do this in OnSize. This is one of the first messages sent to a window, sent after WM_CREATE and before child-windows are created.
So, you might want to try to do it there anyway - it's not much use trying to access controls before they are created, as you've noticed 
Beware, controls might not, for reasons that should be obvious, be created even in OnSize. You can check the m_hWnd-member of them to see if this is the case.
|
|
|
|
|
Thanks a lot. Your last comment was the most helpful. I guess I didn't realize the order in which the window initialization functions are called and I simply needed to check for the control's creation. Thanks again!
|
|
|
|
|
Scolex wrote:
I cannot access these functions for the controls in the OnCreateClient function.
That's because OnCreate() calls OnCreateClient() indirectly before the controls get created.
In the OnCreateClient() method, set a member variable to true indicating that the splitter window has been created. In the OnSize() method, after calling the base class, check to see if the splitter window has been created. If it has, get the height/width of the toolbar and the status bar and subtract both of those from the height/width of the client area. Then for each row/column call SetRowInfo()/SetColumnInfo() using half of the width/height for the second parameter, followed by a call to RecalcLayout(). Make sense?
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
