|
Are there any STL data structure which behave like CMapPtrToPtr
thanks in advance
Sonork 100.41263:Anthony_Yio
|
|
|
|
|
|
|
We have been having very random crashes from an application in production at a customers site. To cut a very long story short Microsoft Australia have told me that it is the result of non thread safe code in the STL shipped with VS6.
We have the following options
1) Upgrade Dinkumware STL from dinkumware site. We would then have to build a custom VC runtime dll with the NEW STL code and ship that as well. Plus the web site includes the following comment
...Please note that this implementation is still not as thread safe as it should be, because of the reference-counted implementation...
2) Upgrade to VS7 dot net. This is Microsofts recommendation.
3) Use Another STL implementation like StlPort
4) Recode system not to use STL
Anyone got any advice? had a similar experiance?
Regards
Josh
|
|
|
|
|
Josh Gray wrote:
Anyone got any advice?
I think the easiest part is option 3). Option 4) is a bit drastic and would require lots of coding (not your idea of a quick fix I imagine).
2) may be easy, but it may also be your worst nightmare. It all depends on if your code compiles and run correctly using that compiler.
Option 5) could be to augment your code with synchronization mechanisms external to the containers. Might be a lot of work though, or it might not. It really depends on your design.
Option 1) is really not an option since, as you say, it has MT issues.
--
Talk to the hand!
|
|
|
|
|
My advice is to use STLPORT and make sure that you compile and use the STLPORT the IOSTREAMS.
John
|
|
|
|
|
We have decided to go the Dot Net route.
Running a build now, only about 1000 signed / unsigned mismatch errors so far and counting!
|
|
|
|
|
Hello Everyone,
I need your sincere help in finding the suitable header file
for the wsnprintf() method. For finding the Domain Users in
Microsoft Active Directory Services(ADSI), i called the ADSI API i.e. FindUsers() to fetch the list of user objects
from a container or directory tree. For that the FindUser()
API code has used the method i.e. wsnprintf() for adding the filter.
RESULT FindUsers(IDirectorySearch *pContainerToSearch, // IDirectorySearch pointer to the container to search.
LPOLESTR szFilter, // Filter for finding specific users LPOLESTR *pszPropertiesToReturn, // Properties to return user objects found.
BOOL bIsVerbose)
This FindUsers() is the API where the definition of the ADSI API calls the wsnprintf() method like as below : DWORD
#define MAX_PATH 520
dwLength = MAX_PATH*2;
// Create search filter.
LPOLESTR pszSearchFilter = new OLECHAR[dwLength];
// Add the filter.
wsnprintf(pszSearchFilter, dwLength, L"(&(objectClass=user)(objectCategory=person)%s)",szFilter);
Please help me out in giving me the details about the header file and the .lib file to make the code compilable.
Thanks,
Derik;)
|
|
|
|
|
I've worked with Microsoft Office development , now I have next problem .
I 've created new Word XP document - E-mail message and try to set programmicaly value of suc fields as Subject,Recepient also add File Attachment . But I can't get access for this fields . I've found that this fields are included in the command bar Envelope that has type msoControlPane . How can I know that class from Microsoft Word Type library 10.0 correspond for this Pane control (_CommandBarButton,_CommandBarActiveX,... or something else) and how can I get access for fields that included in this Pane . Also how can I add extended fields as File Attachment or something else .
|
|
|
|
|
I am using a vector to hold a list of pointers that are newed in the start element handler of the Expat parser. Everything works fine until I exit the application and run my cleanup function.
The following code appears to fail or at least produces a memory leak.
Button *pButton; // pure virtual
vector<button*> vButtons; // holds list of created buttons
// from start handler
pButton = new Label( c_Label );
vButtons.push_back( pButton );
// Cleanup
if( count > 0 )
{
vector<button*>::iterator itClean;
for (itClean = vButtons.begin(); itClean != vButtons.end(); ++itClean)
{
delete *itClean; // crashes here everytime
}
vButtons.clear();
}
Any idea why this is leaking?
-Steve
|
|
|
|
|
What you're doing looks absolutely correct to me. What error do you get when it crashes ?
BTW, this would be easier to read if you checked 'Do not treat <'s as HTML tags' or typed < for <
Christian
I have drunk the cool-aid and found it wan and bitter. - Chris Maunder
|
|
|
|
|
Sorry about the formating, I'll remember next time.
I am not sure whats happening. The destructor for Label is called and then goes into the following
code. I am just getting user break point called and it seems to crash here:
"Label::`scalar deleting destructor'(unsigned int 1) + 37 bytes"
What is a scalar destructor?
004028B5 mov eax,dword ptr [ebp+8]
004028B8 and eax,1
004028BB test eax,eax
004028BD je Label::`scalar deleting destructor'+3Bh (004028cb)
004028BF mov ecx,dword ptr [ebp-4]
004028C2 push ecx
004028C3 call operator delete (0041aec0) // crashes here
004028C8 add esp,4
004028CB mov eax,dword ptr [ebp-4]
004028CE pop edi
004028CF pop esi
004028D0 pop ebx
004028D1 add esp,44h
004028D4 cmp ebp,esp
004028D6 call _chkesp (0041b29e)
004028DB mov esp,ebp
004028DD pop ebp
Sorry, I can't tell you much more that this so far.
-Steve
|
|
|
|
|
smesser wrote:
What is a scalar destructor?
I'd be looking on the object you're deleting to see if it has it's own method you call instead to clean it up. Then I'd try doing nothing and see if memory still leaked ( although it certainly should ).
Christian
I have drunk the cool-aid and found it wan and bitter. - Chris Maunder
|
|
|
|
|
Christian,
I have tryed these sugguestions already. I can't image what is happening. This code used to work.
When it worked I was creating all my buttons in OnInitDialog() and then upon exiting I call
void CHTPCCOREDlg::Cleanup()
{
if ( m_pMemGraphics != NULL )
{
delete m_pMemGraphics;
m_pMemGraphics = NULL;
}
if ( m_pMemBitmap != NULL )
{
::delete m_pMemBitmap;
m_pMemBitmap = NULL;
}
if ( m_pCachedBitmap != NULL)
{
::delete m_pCachedBitmap;
m_pCachedBitmap = NULL;
}
KillTimer( UPDATETIME );
KillTimer( CHECKKEYS );
KillTimer( MOUSEMOVE );
int count = vButtons.size();
if( count > 0 )
{
std::vector<Button*>::iterator itClean;
for (itClean = vButtons.begin(); itClean != vButtons.end(); ++itClean)
{
delete *itClean;
}
vButtons.clear();
}
Gdiplus::GdiplusShutdown(m_gdiplusToken);
}
It was all working fine and then I decided to use XML to allow the buttons to be user definable.
In stead of creating my buttons in OnInitDialog() I am now creating them in the start handler while parsing the configuration file. That should be the only difference and now the code leaks.
Any more suggestions?
-Steve
|
|
|
|
|
Christian,
After some further investigation I am getting this:
memory check error at 0x003267A9 = 0x00, should be 0xFD.
memory check error at 0x003267AA = 0xF0, should be 0xFD.
memory check error at 0x003267AB = 0x41, should be 0xFD.
First-chance exception in HTPCCORE.exe: 0xC0000005: Access Violation.
I have never seen memory errors like this before.
I am wondering if expat is calling its handler functions in another thread and that is the problem. I uncommented my old code and it works perfectly but as soon as I put back in the new code I get the above errors. By the way I did find a couple of other small memory leaks that my new code caused and I am down to the errors mentioned above.
-Steve
|
|
|
|
|
These errors probably indicate you're walking over memory you don't own - I believe debug mode fills a buffer with FD either side of any memory you allocate, so it can tell when you trounce it.
Christian
I have drunk the cool-aid and found it wan and bitter. - Chris Maunder
|
|
|
|
|
If that is the case I can't imagine how since the only difference in the code is where the memory is allocated. This is a strange one.
Thanks for your input.
|
|
|
|
|
This you proably took care of, but is the dtor of Button declared as virtual? Anyway, if it wasn't, the problem would show also with your old code, so I'd look next to the different thread hypothesis.
Joaquín M López Muñoz
Telefónica, Investigación y Desarrollo
|
|
|
|
|
Joaquin,
It ended up being something really strange.
In my class definition I was adding variables as needed.
int ButtonNumber;
CString TitleCaption, SubCaption, DllName,CommandName, Type;
CString FontName, FontStyle, Purpose;
REAL X,Y,W,H;
float TextPad,ImagePad;
Region *region;
Color *BaseColor, *DefaultColor, *FillColor;;
Image *image;
UINT Width, Height;
REAL FontSize;
I don't know why it worked in the old code but in the new code the addition of "REAL FontSize" was causing the crash. I added FontSize to "REAL X,Y,W,H,FontSize" and removed it from the end of the list of variables and it worked correctly.
It seems that the compiler was having problems lining up the memory for the variables if I didn't keep like variables together.
-Steve
|
|
|
|
|
Ummm... this smells very bad. I don't think the compiler is doing a mess with members of your class, most likely you're just masking the error.
Maybe you can try the following: add a trace message in your startHandler outputting the sizeof of the class, and compare with what you've got in the main program. Maybe this sheds some light on the problem.
Joaquín M López Muñoz
Telefónica, Investigación y Desarrollo
|
|
|
|
|
Very good sugguestion. Yes, I know very suspicious behavior.
// crashes
CString Title, Alignment, m_Type;
CString FontName, FontStyle;
int ButtonNumber;
int Padding;
REAL X,Y,W,H;
Region *region;
Color *FontColor;
Color *FillColor;
bool bFill;
REAL FontSize;
// Works fine
CString Title, Alignment, m_Type;
CString FontName, FontStyle;
int ButtonNumber;
int Padding;
REAL X,Y,W,H;
REAL FontSize;
Region *region;
Color *FontColor;
Color *FillColor;
bool bFill;
In all cases and places the size of label is the same (65).
Any other ideas?
|
|
|
|
|
Try using a list instead of a vector to holder the pointers.
|
|
|
|
|
I could try this do you have any reason or explaination to why this would make a difference?
Thanks,
-Steve
|
|
|
|
|
I know why you're crashing!! It happened to me too, and before I discovered why, it was driving me crazy.
Here's the reason (seriously)!! You're trying to delete pointers to memory that have already been deleted.
I discovered that the compiler will call its own destructor on the vector, which will automatically do the cleaning up for you. But later, when you try doing the cleaning up yourself, you are in effect deleting memory that has already been deleted, and that's what causing the crash. IOW, DON'T do the deleting yourself. Just don't do it. The compiler has already done it for you. (I'm talking from experience. It happened to me also.)
William
Fortes in fide et opere!
|
|
|
|
|
William,
Thanks for the insight urr ( experience from the school of hard knocks ).
-Steve
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
