Andrew Brock wrote:

char *p = NULL; *p = 0; //Write 0 to an invalid address

also this does not work. It does nothing, but after some time standard windows crashed program dialog opens and offers me to send crash data to microsoft.

Thank you for tip.

while (!IsDebuggerPresent()) {
Sleep(1000);
}
__asm int 3;

works, but now I have to attach debugger to correct process manually :( It would me more convenient if it would work like before so other ideas are still welcome :)

the same result :(

My application (32 bit) starts next process instance of itself. In new process I need to start debugging so I'm using __asm int 3; It worked good in WinXP/VS2008, it stopped the app and gave me selection whether start new VS instance or use existing VS instance to debug. I've switched to Win 7 64 bit and it does not work anymore. It offers nothing, app just hangs for some time, then exits, I cannot even attach debugger to the process manually. How can I solve this? Thank you. In VS options, the "Jst-in-time" I have all options active (managed,native,script)

modified on Friday, January 21, 2011 5:46 AM

I know what am I doing. I've asked some question. If you don't know the answer, ok. If you have some ideas in area I haven't asked about, I can discuss about it, maybe we both will learn something. But why are you so offensive and writing me to stop posting? You don't know all the details about used algorithm (because it is not important to get answer to my question) but you know I'm complete fool :) If you are concerned about details I haven't wrote, just ask. FYI data is transmitted over HTTPS. The AES is used to encrypt the data. I don't know script kiddies which can crack HTTPS, then AES. Whole point is to make server unable to decrypt the data. Yes server could modify the JS to send the password back, but JS is easy to read and paranoid user can see what is it doing, or can save the HTML+JS locally and use always trusted copy. If you can see some way how can script kid steal other user's data, let me know.

As I've wrote before, I'm not solving authentication problem.

I'm afraid you don't understand what I need to do. I'll try explain in other words. I'm not solving the problem with authentication. User creates some data at his computer. Then encrypts them LOCALLY with some password and upload to server. Server does not know the password, server is not able to decrypt the data! (When somebody hacks the server, he can get the data but has no password). Then user goes to some URL with web application, downloads the data from server (don't think about authentication it is not important now). It will provide pass to JS which decrypts the data. I need to offer storing the password somewhere so he don't need to type the pass each time he opens the application in browser. If he is paranoid then he won't save the password. I was hoping somebody will know whether it is possible to force browser (from JS) to save the password into it's auto-form-filler store which fills the forms.

Mark Nischalke wrote:

Then how do you indend to authenticate the users?

It is not important in my question, it can be done by classic login form (with another pass) + cookies, or HTTP auth, or none. Let's say I've already received encrypted data from server.

Mark Nischalke wrote:

You are a complete FOOL!

Thank you, but I really think it is not possible to save password "safely" in javascript when USER decides to save it. Only to save it encrypted with some master password which would user need to type each time which does not give sense to save the first password in this case.

Server is just storage for client's encrypted data, server must not know the password. I don't worry about secrity of client's password in his browser, it is his problem to secure his computer (actually it is designed primarily for smartphones).

I'm developing HTML application which reads some encrypted data from server, then decrypts it by password provided by user. How can I force browser to save the password in javascript? The password must never be sent to the server, so I cannot use cookies. I know there is some local storage in HTML5, but I'm afraid it is not widely supported. Any ideas? Thank you.

Is there some API to get list of all sockets opened in my process (opened by 3rd party module) and get some usage statistics? (at least last data send/receive timestamp to detect active connections). Thank you.

Thank you, it works, but I think it always creates temporary object so it will be bit inefficient but maybe it won't be noticeable problem.

I have simple template method in my class. I think it is self-explaining:

template<class T> T\* GetCell(int row,int col)
{
	return ((T\*)GridCellGet(col,row,0,0,RUNTIME\_CLASS(T)));
}	

It works in debug build (MFC DLL), but it does not compile in release (static) build. It seems RUNTIME_CLASS() has different definition for DLL and static builds. Error message is: error C2039: 'classT' : is not a member of .... It teems preprocessor replaces RUNTIME_CLASS(T) with 'T' instead of type name, I guess it has higher priority. Do you have idea how to solve this? Thank you.

Interesting point of view, but it doesn't seems to me like exact analogy. When you buy for WIN, you pay for product, which should meet some criteria fixed price, not by hours somebody spent working on it.

Well there is no unit testing and I guess creating some would be 100x more expensive than later fixing of bugs. I haven't seen the code for a year, some parts for 8 years. So it's never 100% safe to touch it. Of course I can find some bugs right after code change and fix it. But some could be found by customers later. When I think of it, it just changes the moment when bug is discovered, so fixing should be paid. Or does/should somebody say to his boss, that today I've broken some function when working at another, don't pay me half day I've spent searching for problem?

So would you fix bugs for free? In bad case it means you make new function and get paid for one hour, but later you can be requested to fix bug you've introduced into old code and it can take 5 hours :doh:

The prove is that thing works in old version. Currently I don't make support, last version is pretty stable, but new functions are needed.

Put in or paper, or not, is different question, first I should know, WHAT to put on it :) Question of bugs introduced into old code is the problem now :)

To be more specific, majority owner is my friend and my partner in another company. Nobody wants to cheat anybody, but we want to make clear agreement to prevent problems later. So it is little bit like moral question. He is not programmer and looks at it like when he calls somebody to fix his car and mechanic breaks something it is his fault and should fix or pay it. For me it is not like that, if I would be employee, I would fix it in my working time as any other task.

I should program some new functions externally into app I've made in previous job. I've never did this before so I don't know what to do if there will occur new bug in old functions because of adding new features? It could occur also later when some user finds it. Is it standard to fix it for free, or charge usual fee per hour? Thank you.