Very Computer

Hi All,

      I have a class that tries to allocate memory dynamically after
doing some processing in a member variable. The class contains:

2 integers,
2 char[256];
and 1 char* ptr;

     This "ptr" would be allocated later on when I get hold of a valid
input.

During initial Creation of the objects, things go through fine. I am
able to allocate memory dynamically for this class. sizeof(myclass)
turns out to be around 1040 bytes.

But after 2 runs, the third time when I try to allocate a new instance
by calling "new myclass()" - it just fails with a Segmentation. The same
code goes through fine the first 2 times. I am afraid I might have
corrupted the malloc/new memory management scheme.  If I actually remove
the "ptr" as a member variable of the class, it works fine any number of
times.  I tried to allocate using malloc which also failed at the same
point.

  cout << "Size of mapping is " << sizeof(mapping) << endl;
  void* ptr1 = malloc(sizeof(mapping));                    ------> fails

  //void* ptr1 = valloc(1040);
------> fails
  cout << " allocated memory\n";
  cout.flush();
  ptr = new mapping(resourceName, 0, this->count);   ------> fails

Can somebody provide help on this?

thanks in advance,
Sabha

Problems like this can be damnably hard to find, because the
cause of the problem can seem to be completely unrelated to
the point where the symptom occurs.  This is why programs
like Electric Fence, or Rational's Purify are popular.  I
seriously doubt that the existence of your pointer is the
cause.  It is more likely something that you are doing
somewhere else that is manifesting when you create your pointer.

--
                Fletcher Glenn

Quote:>   //void* ptr1 = valloc(1040);
> ------> fails
>   cout << " allocated memory\n";
>   cout.flush();
>   ptr = new mapping(resourceName, 0, this->count);   ------> fails

This is damned little to work on.  Can you post more?  Can you post the
whole thing?  Have you tries 'gdb'?

Hi,

       Running with dbx gave the following info:

Woah...  does that even compile on your box?  What system are you using,
anyway?

I can find at least one error right off the bat:

mapping.cpp:
O_RDONLY is not defined.  Need to include fcntl.h

MyTest.cpp: line 1:
#include<HashMap.h>

should be:

#include "HashMap.h"

or else it will look in the wrong directories.  Unless you have your
system set up to look in the current directory for libraries, that won't
work.  And don't expect all over people to be set up the same way (I'm
sure not).

Don't you have a Makefile or something?  Or do you compile all the
modules by hand?

What compiler are you using?

And get a different editor :P  The one you have looks like MS-DOS edit;
there are no tabs, just arrays of spaces.

As for the problem you're having, what is the 'ptr1' variable supposed
to do, anyway?  You malloc it, but then don't do anything with it?  That
just makes SegFaulting more mysterious.

And I'm not familiar, I'm sorry, with dbx.  So the output means nothing
to me.

Yeah, I know that wasn't very helpful, but that's as much as I can tell
you at the moment.

Solaris 2.7

Quote:

> I can find at least one error right off the bat:

> mapping.cpp:
> O_RDONLY is not defined.  Need to include fcntl.h

Might have got removed when I edited some files.

Quote:

> MyTest.cpp: line 1:
> #include<HashMap.h>

> should be:

> #include "HashMap.h"

> or else it will look in the wrong directories.  Unless you have your
> system set up to look in the current directory for libraries, that won't
> work.  And don't expect all over people to be set up the same way (I'm
> sure not).

Sorry - I had everything header file copied into a common location and
included in INCLUDE.

Quote:

> Don't you have a Makefile or something?  Or do you compile all the
> modules by hand?

I do use a Makefile - but it contains parts that are not associated with
this problem - so didnt sent it over - sorry again  :-(

Quote:

> What compiler are you using?

using CC.

Quote:

> And get a different editor :P  The one you have looks like MS-DOS edit;
> there are no tabs, just arrays of spaces.

I was using just plain vi ;-)

Quote:

> As for the problem you're having, what is the 'ptr1' variable supposed
> to do, anyway?  You malloc it, but then don't do anything with it?  That
> just makes SegFaulting more mysterious.

I just wanted to see whether there was problem with new or malloc. new()
kept failing. Then I tried with malloc - just hoping to see it allocate
memory - but it also fails. Anway, you can comment the malloc and only call
new() - it would still fail.

thanks,
Sabha

I second the recommendation to use Electric Fence (or Purify if you've
got it). I've occasionally had problems like this where I (or somebody
else) had pointer problems somewhere else in the code, but the
segmentation fault didn't show up until some random new or malloc.

Steve