|
Hi,
I am currently working on a legacy MultiThreaded MFC application which talks to the remote hardware over a network. The application is compiled with VC++ 6.0 SP5 and has many dlls (Regular, Extension, C++ and WIN32). Now here is a typical user scenarion
1. The user starts a wizard to generate a report for a particular parameter usage on the remote hardware.
2. Configures the report parameters through wizard walkthrough and creates a task that will send some commands to the remote hardware and fetch the required data. The task is executed immidiately.
Now here is a problem I faced during the recent Release build for internal testing due to accidental change in compiler settings.
The application was originally configured to compile with optimization set to "complie for speed" for the Release builds.
When the binary created with this setting is executed, the application runs perfectly fine. The user starts the wizard configure parameters create and execute the task. The application fetches the required data and generates the report.
Now if the compiler setting is changed to optimization=disable with rest every thing remaining same, the application crashes with bad memory reference during the task execution.
During a task execution there may be 3 to 5 threads (worker + GUI) running depending on the complexity of the requested task.
I did found out that the crash was due to invalid thread handle value for one of the threads by attaching the VC debugger externally (On running the release build of the application from within the VC debugger, it works perfectly fine, strange !!!???). Breakpoints are not helping as it seems that they introduces enough time delay for the thread handle to be valid.
I am not too sure on whether the handle is getting invalid either due to not getting initialized at all (the thread is created externally and then set as a reference in class object) or getting terminated too soon.
Well guys I really need help on this as I have been struggling with it for 2-3 days and it is driving me nuts !!
1. I will really like to know what caused the problem at the first place ?
2. What really happened when optimization is turned off that lead to application crash.
I tried to solve this problem by
1. Introducing external delays using loops which didn't help
2. Making the thread switch context by introducing Sleep(0) which did seems to solve the problem (Why?) for this wizard but it crashed another one (again a mistery to me !!)
3. By using Event synchronization object. I create a auto reset event with initial state set as non-signalled. After the thread internal parameters are initialized and before execute method is called I call SetEvent() to set the state to signalled, the application is working perfectly fine with optimization disabled. Interestingly if I simply turnon the optimization for speed without making any code changes everything works fine again !!
In the main execute method I wait for the event to get signalled before proceeding by using WaitForSingleObject(EventHandle, 500), I do get out of the wait but not due to event getting signalled but due to WaitTimeOut happening (don't know why the timeout is happening when the wait should have been over due to event signalling), INFINITE timeout hangs the application.
So please if anyone has a clue on what is going on here please let me know asap.
Sorry if the problem description is bit too long.
Thanks in Advance.
Regards
|
|
|
|
|
Hi,
it seems to me, that you have some weird problem with thread synchronization somewhere. By changing the optimalization switch, the compiler changed the order/amount of instructions that can changed the order/timing of some operation.
This can lead to the situation, where with speed optimalization the thread initialization finishes before other thread uses it, but with optimalization off the initialization is not finished yet, so the second thread uses something uninitialized.
By introducing the Sleep(0) you changed the order of the instructions stream, so it could 'fix' your issue at the moment. But if the synchronization problem is in your application, it will pop-up somewhere else, so definitielly Sleep(0) is not a conceptual solution for such a problem.
By using the event synchronization, you can fix the problem usually (or a part of it). But from the description you gave, it seems that effectivelly you only introduced the 500ms sleep to the start of the application, so just postponing the problem. Only idea I can get is that either the event is not signalled at all, or you can use two different handles, not the same event for synchronization.
hope this helps, dont't hesitate to ask for more
|
|
|
|
|
My C++ application makes system calls, that use cmd.exe ( net use, copy, run .exe, getmac ) with parameters. I am writing a wrapper function that will handle all the parameter formatting, waiting for command to complete, and check the return values.
At first I used the system() command, but I want to hide the cmd window.
So I tried WinExec(), but this runs as a seperate thread and i do see how to suspend the parent thread.
So I tried CreateProcess() but I my exe could be running on any windows platform including Win 95/98. CreateProcess has some compatiblity issues with these version and hiding the window, from what i have read. I am guessing i would have the same problem waiting for the command to complete anyways.
So I tried ShellExecute() but again I was unable to get the parent thread to wait for the command to terminate.
So now I am working with ShellExecuteEx() but again i am having trouble getting the parent thread to wait until the command ends.
can show me how to wait for WinExec, ShellExecute, ShellExecuteEx to finish?
I've been working with WaitForSingleObject but i am missing something..b/c i cannot get it to work.
|
|
|
|
|
jet915 wrote:
My C++ application makes system calls, that use cmd.exe ( net use, copy, run .exe, getmac )
Why use this antiquated approach when functions exist for just such purposes?
jet915 wrote:
So I tried CreateProcess() but I my exe could be running on any windows platform including Win 95/98. CreateProcess has some compatiblity issues with these version and hiding the window, from what i have read.
It works on Windows 95 according to MSDN. What specifically did you read?
jet915 wrote:
So I tried ShellExecute() but again I was unable to get the parent thread to wait for the command to terminate.
It's part of the PROCESS_INFORMATION structure.
jet915 wrote:
can show me how to wait...
PROCESS_INFORMATION pi;
...
CreateProcess(..., &pi);
WaitForSingleObject(pi.hProcess, INFINITE);
"One must learn from the bite of the fire to leave it alone." - Native American Proverb
|
|
|
|
|
|
hi, I download the windbg and tried to use to debug my code.
but how to attach the project with the windbg? seems like VS and windbg is running seperate, does anybody has step by step documentation? thank you!
I can't find any moeory.dum file in my machine.
|
|
|
|
|
Check out Toby Opferman's article(s).
"One must learn from the bite of the fire to leave it alone." - Native American Proverb
|
|
|
|
|
Thanks for reply. I am looking for the memory dump file in my machine, but no luck so far, wondering if the windows2000 could not generate such file....?
thank you for any info!
|
|
|
|
|
If a control has tabstop set to true then it can also be accessed with the mnemonic access keys. I have a control that has tabstop set to false (don't want user to be able to tab to it) but do want to allow access via the access key. I've noticed that it doesn't work when tabstop set to false. I've even tried setting a static control with tab order preceding the control to help get the caret there but no luck. Would I have to intercept a windows message to do this? Saw in Spy++ some GetDlgCode messages being fired during the Access Key event but not sure what to do. Thanks
|
|
|
|
|
mx483 wrote:
...the access key.
Do you mean mnemonic?
"One must learn from the bite of the fire to leave it alone." - Native American Proverb
|
|
|
|
|
Yep mnemonic. I even tried an accelerator key that was mapped to code that would set the focus but I haven't gotten that to work either. Back to the drawing board.
|
|
|
|
|
Was successful in mapping an accelerator key to some code that put the focus to a control that had tabstop set to false. It wasn't working originally because I was using a dialog app that probably didn't have the accelerator table attached. Curious if there are other solutions. This seems to work pretty well though.
|
|
|
|
|
i dont see menu on this code . can some1 help me . how can i view a pop-up menu programaticaly
m_menu.m_hMenu = NULL;
m_menu.CreatePopupMenu();
CMenu childmenu;
VERIFY(childmenu.CreatePopupMenu());
m_menu.AppendMenu(MF_STRING | MF_POPUP | MF_ENABLED, (UINT)childmenu.m_hMenu, _T("&Delete"));
childmenu.Detach();
childmenu.DestroyMenu();
///BOOL res1 = m_menu.EnableMenuItem((UINT)childmenu.m_hMenu, MF_ENABLED);
BOOL res2 = SetMenu(&m_menu);
///this->DrawMenuBar();
|
|
|
|
|
I don't really see what you want to do but if you want to show a pop-up menu, you will need to call TrackPopupMenu member function from CMenu.
|
|
|
|
|
|
i'm pushing values inside a vector using push_back()...
then.. to print it can't we use cout< >::*)(void) const'
Error executing cl.exe.
Thanx
V
|
|
|
|
|
That would help a lot if you were using the < tag that is just above the emoticons (your code is impossible to read, and the same for the error message). And also, put the declaration of the vector itself.
|
|
|
|
|
#include <vector>
using namespace std
{
vector<int> vec;
vec.push_back(1);
vec.push_back(2);
vec.push_back(3);
vec.push_back(4);
for(int i=1;i<vec.size;i++)
cout << vec[i];
}
thanks,
V
|
|
|
|
|
The problem is not on cout, the problem is at this line:
Vivekuniq wrote:
for(int i=1;i<vec.size;i++)
You forgot the parenthesis after size. Try this instead:
for(int i=1;i<vec.size();i++)<br />
|
|
|
|
|
It seems that you are trying to write the data where you should evalute the lopp condition ... but if you could write the whole piece of code would be quite more understandable  ...
Marc Soleda.
... she said you are the perfect stranger she said baby let's keep it like this... Tunnel of Love, Dire Straits.
|
|
|
|
|
marcdev wrote:
It seems that you are trying to write the data where you should evalute the lopp condition
No, it's just because the '<' symbol is interpreted as an HTML tag so a part of his code has been removed.
|
|
|
|
|
Ok, i didn't realise about that !!! 
... she said you are the perfect stranger she said baby let's keep it like this... Tunnel of Love, Dire Straits.
|
|
|
|
|
|
#include <vector>
using namespace std
{
vector<int> vec;
vec.push_back(1);
vec.push_back(2);
vec.push_back(3);
vec.push_back(4);
for(int i=1;i<vec.size;i++)
cout << vec[i];
}
Thanx
V
|
|
|
|
|
sh*t, this is silly . sorry guys...
Thanks a lot
V
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
