C / C++ / MFC

Re: I tried PostMessage as well

2-Feb-17 16:17

I was told PostMessage would work for IPC the child process never gets control

Victor Nijegorodov2-Feb-17 22:03

Victor Nijegorodov

2-Feb-17 22:03

How do you use it?
What message ID do you use?

Re: I tried PostMessage as well

leon de boer2-Feb-17 23:48

2-Feb-17 23:48

The child probably wont see the message because the parent isn't passing it up. Very few messages are passed from a parent to a child unless you specify it beyond the obvious stuff like keyboard and mouse. Depending on the ID it isn't clear to me a message will be passed to a child window I need to know more about the window types and flags involved.

In vino veritas

modified 3-Feb-17 5:59am.

IPC Thru WM_COPYDATA

ForNow2-Feb-17 15:52

2-Feb-17 15:52

Hi
I have been having intermittent problems notifying a child window process with SendMessage

So I decided to go with the Microsoft Sanctioned way WM_COPYDATA

In my console Parent process I do GetConsoleWindow (which I observe returns what seems

a valid HWND),

I declare a COPYDATASTRUCT member on the stack (locally)

and do SendMessage I have gotten the child window HWND with FindWindow

in my MFC child window I have a messagemap entry

and nothing it never gets there I am running under both the console app and the MFC app

under Visual studio debugger so I can see what's going on

C++

ON_MESSAGE(WM_COPYDATA,Debug_it)

Jochen Arndt2-Feb-17 21:36

Jochen Arndt

2-Feb-17 21:36

See the first comment at WM_COPYDATA message (Windows)[^]:

Quote:
Starting with Windows Vista, the application that is to receive the WM_COPYDATA message must call the ChangeWindowMessageFilterEx function. Otherwise the the message will not be received. It will be blocked by UIPI.

ForNow4-Feb-17 16:46

4-Feb-17 16:46

Thanks for your help I just tried

C++

ChangeWindowMessageFilterEx

I got a FALSE GetLastError returned ERROR_ACCESS_DENIED

This is because my process is at SECURITY_MANDATORY_LOW_RID

AnyWay I can Change this. Thing is I understand the concern for security however I (my process ) created the child process so I should be able and send messages

Jochen Arndt4-Feb-17 23:37

Jochen Arndt

4-Feb-17 23:37

It's not a problem of sending messages but of receiving. So the only solution would be increasing the UIPI level of your MFC application or using another IPC method.

ForNow5-Feb-17 3:01

5-Feb-17 3:01

Thanks I didn't read it entirely
For a Mainframe in the OnCreate and for a modless dialog I'll put it in the OnInitDialog

ForNow5-Feb-17 14:19

5-Feb-17 14:19

Still works intermittently I am going to try DDE since my code is based on messages

Rick York9-Feb-17 6:58

Rick York

9-Feb-17 6:58

Many, many years ago I wrote a very simple app to demonstrate the use of WM_COPYDATA with MFC and I see that it is still online : Inter-Process Communication Using WM_COPYDATA[^]

I rebuilt it just now in VS2015 (with very little effort) and it still works on W10. You might want to have a look at it as an example.

ForNow9-Feb-17 8:22

9-Feb-17 8:22

I originally had the WM_COPYDATA as a ON_MESSAGE message map handler than I noticed the OnCopyData I wasn't able to get it to work with that as well

My current method works fine as I don't have to wait for a reply with the SetEvent and shared storage thanks for your help

leon de boer2-Feb-17 23:45

Thr problem of the recursive algorithm to do binary search in a sorted array

2-Feb-17 23:45

I am not sure the copydata can be local but for safety I would just use globalalloc anyhow

HGLOBAL hgbl = GlobalAlloc(GMEM_ZEROINIT, sizeof(COPYDATASTRUCT) );  // Allocate the global memory
 COPYDATASTRUCT* cpd = (COPYDATASTRUCT*)GlobalLock(hgbl);        // Lock the allocated memory to a pointer

//..... use an abuse the pointer .. when you are done

GlobalUnlock(hgbl);             // Release lock on the memory block

// Reciever then locks and plays and unlocks

// ......zzzz sometime later one of the two parties

GlobalFree(hgbl);	            // Free the allocated memory

Not sure it's the problem but at least then when know the memory passed is valid to both parts.

In vino veritas

modified 3-Feb-17 6:06am.

alalba2-Feb-17 15:37

Re: Thr problem of the recursive algorithm to do binary search in a sorted array

2-Feb-17 15:37

I want to inplement a binary search algorithm of an order array.

int Array::Search(int target, int* start, int* tail){
	if(start>tail){
		cout<<"not find"<<endl;
		return -1;
	}
	int offsetMid= (tail-start)/2/sizeof(int);//by unit of int
	cout<<"offset: "<<offsetMid<<"  mid " <<*(start+offsetMid)<<endl;
	cout<<start<<" " <<tail<<endl;
	if(target == *(start+offsetMid) ){
		cout<<"mid"<<endl;
		return (offsetMid+1);
	}
	if(*(start+offsetMid) > target){
		cout<<"left"<<endl;
		tail = start+(offsetMid-1)*sizeof(int);
		return Search(target, start, tail);
	}
	if(*(start+offsetMid) < target){
		cout<<"right"<<endl;
		start = start+(offsetMid)*sizeof(int);
		return Search(target,start , tail);
	}
}

But it doesn't work, I know the parameter of the func should better be the index of the array rather than a pointor, I just want to try this way. The compiler I used is

gcc version 6.2.1 20160916 (Red Hat 6.2.1-2) (GCC)

Another weird thing

int offsetMid= (tail-start)/2/sizeof(int);//by unit of int

Both tail and start are the type of int*, but the difference of them is counted by byte, so I must divide it by sizeof(int) to get the offset value.

if(*(start+offsetMid) > target)

But when I want to plus the offset value to the start address to get the value in a particular position, I shouldn't time it by sizeof(int).WHY??

Midi_Mick2-Feb-17 16:12

Midi_Mick

2-Feb-17 16:12

You have already asked this question in QA. Please do not repeat questions in multiple places. Choose your forum, and stick with it.

Cheers,

Mick
------------------------------------------------
It doesn't matter how often or hard you fall on your arse, eventually you'll roll over and land on your feet.

Re: Thr problem of the recursive algorithm to do binary search in a sorted array

alalba2-Feb-17 16:22

Re: Thr problem of the recursive algorithm to do binary search in a sorted array

2-Feb-17 16:22

I'm so sorry about that. This is my first time come here, I'm not familiar with here.So I just want more people can see my problem. Thank you for you advice.

alalba3-Feb-17 4:05

Re: Thr problem of the recursive algorithm to do binary search in a sorted array

3-Feb-17 4:05

U are totally right, I found I made a silly mistake in Creator function that I DID'T post it up . Thank u for your answer and your patient, I hope my moron question didn't make you confused, if it is, I'm so sorry about that and the time u waste on my question.

leon de boer2-Feb-17 23:18

Re: Thr problem of the recursive algorithm to do binary search in a sorted array

2-Feb-17 23:18

Ok I want you to get out pen an paper and do this on a piece of paper

You have have two integers in an array and the pointer to the first one we will call zero. So the second pointer must be exactly one integer away so it will be sizeof(integer) from the first pointer.

Can you get what the result of this is (tail-start)/2 .. I will give you a hint if an integer is 2 bytes the answer is 1 byte. It doesn't matter what divisions you do after that you have lost alignment to the array your result can be some point in the middle of an integer.

Nothing in all the code above deals with that problem and it's really simple to fix all you need is to add a simple +1 in exactly the right place. Start by fixing the alignment problem and most of the rest will drop out which was sort of the second you cottoned onto what is the middle of the array and what does that actually mean. Take a one integer array what is the middle mean, your code has to know what to do with pathological case. The two integer array is an edge case on dealing with even number arrays where the middle of the array isn't aligned to an integer pointer.

This whole concept is sometimes called the post and fence problem, where you always need one more post than the gaps you span if you need to search.

modified 3-Feb-17 5:26am.

alalba3-Feb-17 4:25

Re: Thr problem of the recursive algorithm to do binary search in a sorted array

3-Feb-17 4:25

thank u for your answer and your patient. You are so nice! It really teach me a lot, I get more than I thought. Thank u!

leon de boer3-Feb-17 7:12

Re: Thr problem of the recursive algorithm to do binary search in a sorted array

3-Feb-17 7:12

No worries can I also give you a much safer option which I always try to teach to new graduates

Don't move the entries at all in the sort. Make yourself an array of pointers to the entries and shuffle your pointers in your private array.

The reasons are obvious
1.) The is no risk of damaging the original data and if you want to undo the function it's instant
2.) On most systems these day an integer is 32bits as is a pointer so it is the same speed to sort them.
3.) For complex types doubles, floats, strings the pointer is a lot smaller than the data.
4.) The source data doesn't have to be in the same place, it can be spread out all over the place the pointers don't care

A common test I set for new employees look like this .. Please write a routine to sort these animals

char* DomesticAnimals[4] = { "Cat", "Dog", "Goat", "Cow" }; // Some domestic animals .. feel free to add some
char* RodentAnimals[4] = { "Mouse", "Rat", "Rabbit", "Fox" };	// Some rodent animals .. feel free to add some
char* WildAnimals[4] = { "Elephant", "Rhino", "Lion", "Tiger"}; // Some wild animals .. feel free to add some,

Basically you copy or move the entries you don't get the job.
The three different source arrays throws most graduates for a loop they feel they have to move something.
Even if they go for an "in place" sort they put wrong animals in wrong categories and they foul my database.

You don't move anything, you make you own private array of char* pointers and point alternate strings and sort whats pointed to with a routine. There are 12 entries above so all you need is

char* myArray[12];

Load each entry to a pointer in that array and sort, and you don't move anything and you don't risk my data if you botch it. This example above sort of makes it very clear why you don't ever move the data .. not ever!

If you want to see the trick of sorting pointers large scale open a windows folder viewer .. what do you think they are doing they don't ever really move or copy the data in anyway it's too risky and slow.

It sorts out those who can program from those who can just write code very quickly.

I would expect a good programmer to be able to write a bubble or quick sort on that array in about 20 lines of code.
Here is the standard Rossetta code for it which is 15 lines
Sorting algorithms/Quicksort - Rosetta Code[^]

In vino veritas

modified 3-Feb-17 13:39pm.

alalba3-Feb-17 12:46