|
Hi,
Can you please explain what is sequence point.
And one more thing if my function is fun(int&1, int&j, int &k) and if i try to call fun(++i, ++i, i++)
i will get error at 3rd parameter.
It is showing problem for i++ only, not for ++i.
|
|
|
|
|
Anonymous wrote:
It is showing problem for i++ only, not for ++i.
It will show the problem for ++i as well. The reason it doesn't show the error is because the compiler is evaluating the argument right-to-left, and is stopping when it finds one in error - the 3rd one in this case.
Anonymous wrote:
Can you please explain what is sequence point.
A sequence point is a point in the code where all the side-effects up to this point have been evaluated. They occur at the end of a statement, at the comma operator, the && and || operators, and when a function is called after all the arguments have been evaluated. It is at these points that the effects of the ++ and -- operators occur. So in your case, you are modifying the value of i twice within a sequence point (the evalution of the function parameters).
Ryan "Punctuality is only a virtue for those who aren't smart enough to think of good excuses for being late" John Nichol "Point Of Impact"
|
|
|
|
|
Hi Ryan,
But each expression one after the other. So in C/C++
Fun(++i, i++) first i++ will be evaluated and then ++i is evaluated and after that only Fun will be called.
But my surprise is why
int i = 10;
printf("%d, %d, %d", ++i, i++, ++i); will print as 12, 11, 11.
But what I anm expecting is 13, 11, 11.
What is the reason?
|
|
|
|
|
Ranjish wrote:
So in C/C++
Fun(++i, i++) first i++ will be evaluated and then ++i is evaluated and after that only Fun will be called.
No. The order of evaluation of function arguments is undefined. It is not specified by the standard, and the compilers can do whatever they like.
The results of modifying the same variable twice (or even three times) in a function call like you're doing is completely undefined. It's totally random. There is no way you can tell what is going to happen. The compiler doesn't even have to generate correct code for this case. In my opinion, the compiler should flag an error, because it's impossible to work out what should happen. Even in your case when you're expecting 13,11,11 the compiler could legally set every one of them to random numbers, because you're trying to do something undefined.
Basically, do not ever change the value of a variable twice in a single statement. The results are completely undefined, and could be anything.
Ryan "Punctuality is only a virtue for those who aren't smart enough to think of good excuses for being late" John Nichol "Point Of Impact"
|
|
|
|
|
Both of these operators return an int, not an int&, so you can only pass them to a function expecting an int or a const int&.
Besides, according to the C++ standard, the values passed to your function will be undefined because you are changing a variable value twice between the same sequence points (in the function call). The values will differ depending on which order the compiler evaluates the parameters in; the order is not defined in the standard and is compiler-dependent.
Ryan "Punctuality is only a virtue for those who aren't smart enough to think of good excuses for being late" John Nichol "Point Of Impact"
|
|
|
|
|
[edit] Just Googling, I got this from this page: http://cpptips.hyperformix.com/cpptips/inline_seq_pt[^]
the result would either be `1 - 2' OR `2 - 1'. But, it is a safe
operation and the result MUST be one of these two values. Please note
this is very different from code like:
int main() {
int x = 0;
return ++x - ++x;
}
which results in UNDEFINED behavior. This program may return -1, 1,
42, or it may crash. That's because it modifies the same variable in
an expression without an intervening "sequence point". There's a
"sequence point" at the call to a function after evaluating all the
arguments (which is why the first two main() functions are legal).
There's also sequence points at the operators &&, ||, ?, and comma.
My articles
www.stillwaterexpress.com
BlackDice
|
|
|
|
|
Because i++ returns an int, not a reference to an int.
But in any event you should never write code like that, because having two side effects working on the same variable in the same line of code leads to undefined behavior.
--Mike--
LINKS~! Ericahist | 1ClickPicGrabber | CP SearchBar v2.0.2 | C++ Forum FAQ | You Are Dumb
Magnae clunes mihi placent, nec possum de hac re mentiri.
|
|
|
|
|
Hi all,
I need to start an automatic procession of any files apearing in a certain folder.
Of course I can work with a counter, to check the folder every once in a while. But I´m sure there is a more elegant way of programming that.
Does any one know, if there is anything hidden in the winapi or elsewhere, that makes windows notify my client in case of added files?
Or does anybody know, that there is no alternative? That would at least help me feel better 
Thanks.
Stefan.
|
|
|
|
|
|
Also check out FindFirstChangeNotification().
"Opinions are neither right nor wrong. I cannot change your opinion. I can, however, change what influences your opinion." - David Crow
|
|
|
|
|
There's an article right here[^] on CP that discusses this.
'til next we type...
HAVE FUN!! -- Jesse
|
|
|
|
|
I am working on VC++, now I want to search Front Page from windows Registry. But the front page path would not be same for each new version, so i can not fix its path. So please tell me how do i search front page from windows registry.
Thanks in advance.
Regards,
Lavani
|
|
|
|
|
What are you searching for? Given that FP can be installed anywhere, there's no static location that you can look in. You'll either need to prompt the user for the location, or search the disk(s) for the file(s) in question.
"Opinions are neither right nor wrong. I cannot change your opinion. I can, however, change what influences your opinion." - David Crow
|
|
|
|
|
Thanks,
Ok, I agree that FrontPage can install anywhere, But in the windows registry it may have a entry for FP. I want to search for windows registry for FP without fixing its path through VC++. Or
Can I Search windows registry for default Web page editor?
If I found default Web Page Editor. Can I get its path from the Windows Registry?
Regards,
Lavani
|
|
|
|
|
Thanks,
Ok, I agree that FrontPage can install anywhere, But in the windows registry it may have a entry for FP. I want to search for windows registry for FP without fixing its path through VC++. Or
Can I Search windows registry for default Web page editor?
If I found default Web Page Editor. Can I get its path from the Windows Registry?
Regards,
Lavani
|
|
|
|
|
lavanii wrote:
But in the windows registry it may have a entry for FP...
Right, but even if one exists, you'd still require an exhaustive search to find it.
lavanii wrote:
Can I Search windows registry for default Web page editor?
Yes, but there are no guarantees. See AssocQueryString() for more.
"Opinions are neither right nor wrong. I cannot change your opinion. I can, however, change what influences your opinion." - David Crow
|
|
|
|
|
|
Rational Purify Plus
http://www-306.ibm.com/software/awdtools/purifyplus/features/index.html
|
|
|
|
|
Hey there...
I'm having a problem with MessageBox(HWND, char*, char*, UINT). I'm writing C++ to the WinAPI, no high level abstractions other than those that I roll (ie no MFC, ATL, or managed extentions).
I have a main window created with WNDCLASSEX. Durig WM_CREATE, a couple dialog boxes are created and displayed to help guide the user through a couple of setup steps. The problem lies in calling MessageBox(...) from any of these dialogs. Any attempt to display a MessageBox(...) results in a system beep, all GUIs lose focus, but the MessageBox never appears.
If I run the debugger, I can watch the system get to the MessageBox(...), execute it, return IDOK, then the debugger quits and lets the application resume, never executing any code that follows.
If I remove the calls to show and update the main window (in WinMain(...)), then the dialogs can display MessageBox(...) as normal. The Dialogs are also able to call one another with no ill effects, regardless of whether or not I show the main window.
Now, I have a template I have written over time, which includes WinMain, WinProc(...), etc, so I just have to cut and paste to get an application started, and none of my previous applications have ever had a problem like this. I have spent days now, comparing this app to previous ones, and I simply cannot find a difference between the architecture and the calls (nor the way I use them).
I have tried everything I can think of, tweaking every dialog attribute there is, changing the calls and the order the are invoked. One other note, if I invoke a MessageBox(...) from within WM_CREATE, all is fine, its only after the ShowWindow() and UpdateWindow() calls that this behaviour starts to exhibit itself.
Has anyone ever seen this behavior before? Have any idea what it is that I could have done wrong this time?
Some code samples follw this post.
Thank you for reading all this and any help you might be able to offer
Dave
p.s. compiling with VC++ 7.1 on Win XP Pro
from WinMain....
// store the instance
DAT->hInst = instance;
MSG msg;
WNDCLASSEX wcex;
HWND hWnd = NULL;
// build the windows struct
wcex.cbSize = sizeof( WNDCLASSEX );
wcex.style = CS_HREDRAW | CS_VREDRAW;
wcex.lpfnWndProc = (WNDPROC)WinProc;
wcex.cbClsExtra = 0;
wcex.cbWndExtra = 0;
wcex.hInstance = DAT->hInst;
wcex.hIcon = LoadIcon( DAT->hInst, (LPCTSTR)ICO_GLOGO );
wcex.hCursor = LoadCursor( NULL, IDC_ARROW );
wcex.hbrBackground = (HBRUSH)GetStockObject( GRAY_BRUSH );
wcex.lpszMenuName = (LPCTSTR)MENU_MAIN;
wcex.lpszClassName = gas::sAPPNAME.c_str();
wcex.hIconSm = LoadIcon( wcex.hInstance, (LPCTSTR)ICO_GLOGO );
if( !RegisterClassEx(&wcex) )
{
MessageBox( NULL, "Error Registering Window: Need NT?", "Application Error", MB_ICONERROR | MB_OK );
}
else
{
hWnd = CreateWindow( gas::sAPPNAME.c_str(), std::string(gas::sAPPNAME + " " + gas::APPVER).c_str(), WS_OVERLAPPEDWINDOW, CW_USEDEFAULT, 0, CW_USEDEFAULT, 0, NULL, NULL, DAT->hInst, NULL );
if( !hWnd )
MessageBox( NULL, "Error Creating Window: NULL Handle Returned!", "Application Error", MB_ICONERROR | MB_OK );
else
gas::Data::hMain = hWnd;
}
// everything is done, give the user the application window
ShowWindow( hWnd, showcmd );
UpdateWindow( hWnd );
// welcome the user if they still wish to be
if( CFG->GetShowWelcomePrompt() && gas::Data::hWelcome ) {
ShowWindow( gas::Data::hWelcome, SW_SHOW );
SetFocus( gas::Data::hWelcome );
}
while( GetMessage( &msg, NULL, 0, 0 ) > 0 )
{
if( !IsDialogMessage( DAT->hSplash, &msg ) &&
!IsDialogMessage( DAT->hDataSrc, &msg ) &&
!IsDialogMessage( DAT->hWelcome, &msg ) )
{
TranslateMessage( &msg );
DispatchMessage( &msg );
}
}
return( static_cast<int>( msg.wParam ) );
from WinProc::WM_CREATE
case WM_CREATE:
{
// gonna do a simplified splash screen for now
// later this is where we'll retreive startup data
// and perform synchronization
gas::CreateSplashDlg( hwnd );
if( gas::Data::hSplash )
{
ShowWindow( gas::Data::hSplash, SW_SHOW );
SendMessage( gas::Data::hSplash, WM_COMMAND, MAKESHIFT_TIMER, 0 );
}
if( CFG->GetShowWelcomePrompt() )
{
gas::CreateWelcomeDlg( hwnd );
if( !gas::Data::hWelcome )
MessageBox( NULL, "Error: Failed to create the Welcome Prompt GUI", "System Error", MB_OK|MB_ICONERROR );
}
return 0;
}
From one of the dialog headers...
case BUT_DATASRC_ACTIVATE_ONE:
{
// which one do they want to activate
unsigned int index = (unsigned int)SendMessage( GetDlgItem(hwnd, LST_DATASRC_DEFINED_SRC), LB_GETCURSEL, 0, 0 );
if( index >= 0 && index < DAT->GetDataSourceCount() )
{
gas::DataSource * ds = DAT->GetDataSource( index );
ds->active = true;
PopulateListBoxes( hwnd );
}
else
{
MessageBox( NULL, "You must select a data source before you can activate it", "Application Error", MB_ICONERROR|MB_OK );
}
processed = TRUE;
break;
}
|
|
|
|
|
I usually NEVER pass in NULL as the first argument. Just for grins, try using the desktop window retreived with the GetDesktopWindow() function and see if that works for you.
|
|
|
|
|
Hey Blake...
Thanks for the reply. I did as you instructed and and expereinced the exact same behavior.
Regarding the HWND arg, I typically pass in the handle of the GUI invoking the MessageBox(), but in an attempt to solve this I was pouring through Petzold, and one of the things he notes is that NULL makes it so that the MessageBox() isn't owned by anyone. I have been thinking I have some sort of message pump issue with the message making its way to the main window, so I thought I'd try eliminating ownership.
Honestly, I am at a loss... I have been comparing this code to half a dozen of apps I've written and I just can't find a difference. I invoke MessageBoxes from dialogs all the time.
Thanks for taking a look and adding some input
Dave
|
|
|
|
|
Generally, invoking the message boxes from dialogs is NOT a problem.
What I am beginning to suspect is that if this current work is done when your application is starting up, you don't have a message pump for your program's primary thread yet. Microsoft's most recent guidelines and documentation infer that a thread does not necessarily have a message pump associated with it. A message pump is not created until the first call that would post, send, or attempt to retreive a message is made. You now might try putting a blank PeekMessage at the beginning of your program, using NULL as the 'window handle' to see if that will help your program get a message pump started. Then the message boxes might work after that call has been made.
A program's primary thread normally gets the message pump started when it first enters the get/translate/dispatch message loop. Anyway, it is just an idea of what I would be trying at this point to make it work.
|
|
|
|
|
Hi Blake...
This was a good thought, I was hopeful. I tried at the very begginging, after the window creation and in the WM_CREATE... nada.
I did however take a cue from you post. After the failed attempt with the peek, I removed all calls from the WM_CREATE, let the main window complete (let it get a WM_SIZE and WM_PAINT). Added a menu and then invoked the dialogs via the menu after the main window had time to process a few messages. The behavior was unchanged.
I've been running diff after diff trying to find where I strayed from past projects and just can't find it. So I am thinking that I should start a new project and just rebuild piece by piece until I find where things go to hell. Instead of my template, I'll copy a project that works and edit it as needed to get it going.
Thanks for all the input, it was GREATLY appreciated!
Dave
|
|
|
|
|
Just to follow up on this issue...
I fixed it, but still havn't solved it.
I created another project with my templates and got that working as normal. I then went back to the original project and commented out the entire main file.
I cut and pasted in the main file from the new project, made sure it still worked, and then proceeded to cut and paste my original code back into place line by line.
Esentially I am back to my original code, byte for byte unchanged, but it works... ??????????
I am 1000% confused, have never seen anything as annoying as this, but I am back in business again which is the most important thing for now.
Thanks for all your help!
Dave
|
|
|
|
|
Hello Friends,
I had been posting my problems for the system beep and finally after alots of experiments I found the person who was responsible for the beep sound. It was the Clear() member function of the CRichEditCtrl class.
Now please can anybody tell why this member function is giving a beep sound (Default Beep) of the system properties.
Thanks in advance.
Neelesh K J Jain.
|
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
