a weird memory problem

[neoromance]

I ran into a weird memory problem.

my c++ code got a segmentation error at the execution. why I try to
debug it with gdb, I found out it was due to an earlier memory problem.
I then used both valgrind and efence tools, both tell me that an early
'new' operation failed.

But I have a try{}catch(std:bad_alloc){} block to catch bad memory
allocation exceptions. why it wasn't caught?

By the way, I have more than enough momory available for the code, why
i still get a bad memory allocation error?

thanks

[Mike Wahler]

<ne********@gmail.com> wrote in message
news:11**********************@o13g2000cwo.googlegr oups.com...

I ran into a weird memory problem.

my c++ code got a segmentation error at the execution. why I try to
debug it with gdb, I found out it was due to an earlier memory problem.
I then used both valgrind and efence tools, both tell me that an early
'new' operation failed.

But I have a try{}catch(std:bad_alloc){} block to catch bad memory
allocation exceptions. why it wasn't caught?

By the way, I have more than enough momory available for the code, why
i still get a bad memory allocation error?

Bug on line 42.

-Mike

[neoromance]

can you explain it in more detail?
thanks

[Jack Saalweachter]

ne********@gmail.com wrote:

can you explain it in more detail?
thanks

On the internet, people with small pipes frequently resort to a
highly-compressed form of communication, in order to save time and money
on uploading their replies. "Bug on line 42." decompresses to "You have
described how woeful your problem is in great detail, and your sorrows
are indeed sorrowful, but you have not adequately /described/ your
problem, perhaps by posting reasonably sized snippets of the code which
is causing you issues. Perhaps if you were to describe your code and
your problem more thoroughly, we might help you, but until then we can
only speculate wildly and fruitlessly."

Obviously, "Bug on line 42." is much more concise.

[neoromance]

:) thanks

here is my problem:

I have 2 classses, node and edge. I construt a graph by constructing
one node or one edge at a time.
I have more than enough memory available.

the code works fine. But when I try to use valgrind to track memory
leak, it gave me a error "can't allocate memory" when the code created
half of the nodes and edges. Same thing happened when I link my code
with efence library.

In my code, I have try{}catch(){} block to test bad allocation. So it
seems without efence wrapper or valgrind, the allocation is fine, and
with them, I got a bad allocation error, and the error was not caught
by my code.

thanks

[Kai-Uwe Bux]

ne********@gmail.com wrote:

can you explain it in more detail?
thanks

Explain what?

When you reply to a post, please quote the relevant parts.
Best

Kai-Uwe Bux

[Mike Wahler]

<ne********@gmail.com> wrote in message
news:11**********************@g49g2000cwa.googlegr oups.com...

:) thanks

here is my problem:

I have 2 classses, node and edge. I construt a graph by constructing
one node or one edge at a time.
I have more than enough memory available.

the code works fine. But when I try to use valgrind to track memory
leak, it gave me a error "can't allocate memory" when the code created
half of the nodes and edges. Same thing happened when I link my code
with efence library.

In my code, I have try{}catch(){} block to test bad allocation. So it
seems without efence wrapper or valgrind, the allocation is fine, and
with them, I got a bad allocation error, and the error was not caught
by my code.

Please read Jack's reply again, this time carefully.

-Mike

[Kai-Uwe Bux]

ne********@gmail.com wrote:

:) thanks

here is my problem:

I have 2 classses, node and edge. I construt a graph by constructing
one node or one edge at a time.
I have more than enough memory available.

the code works fine. But when I try to use valgrind to track memory
leak, it gave me a error "can't allocate memory" when the code created
half of the nodes and edges. Same thing happened when I link my code
with efence library.

In my code, I have try{}catch(){} block to test bad allocation. So it
seems without efence wrapper or valgrind, the allocation is fine, and
with them, I got a bad allocation error, and the error was not caught
by my code.

From this description, it is impossible to say anthing meaningful about the
problem. It could be a bug in valgrind or efence. It could be a bug in your
code. Without seeing what causes the problem, we cannot help you.

Hint: you may want to post some relevant code.
Best

Kai-Uwe Bux

[Karl Heinz Buchegger]

ne********@gmail.com wrote:

I ran into a weird memory problem.

my c++ code got a segmentation error at the execution. why I try to
debug it with gdb, I found out it was due to an earlier memory problem.
I then used both valgrind and efence tools, both tell me that an early
'new' operation failed.

But I have a try{}catch(std:bad_alloc){} block to catch bad memory
allocation exceptions. why it wasn't caught?

It may be, that your real bug happend way before that bad allocation.
Eg. You overflowed an array and thereby destroyed the bookkeeping data
structure of the memory allocator. Or you destryoed the return value
stack (if your system uses a stack). Things like that.

--
Karl Heinz Buchegger
kb******@gascad.at

[DellBoy]

You mention that you have enough memory to run all three apps at once.
It might be the case that the memory itself is at fault. Problem memory
sticks seem to be commonplace these days. Try testing your RAM:
http://www.memtest86.com/#download0

[John Harrison]

ne********@gmail.com wrote:

:) thanks

here is my problem:

I have 2 classses, node and edge. I construt a graph by constructing
one node or one edge at a time.
I have more than enough memory available.

the code works fine. But when I try to use valgrind to track memory
leak, it gave me a error "can't allocate memory" when the code created
half of the nodes and edges. Same thing happened when I link my code
with efence library.

In my code, I have try{}catch(){} block to test bad allocation. So it
seems without efence wrapper or valgrind, the allocation is fine, and
with them, I got a bad allocation error, and the error was not caught
by my code.
It likely that an even earlier bug in your code caused the problem. As
youve already found out bugs often manifest themselves much later than
the actual problem.

Once a bug has corrupted your heap all bets are off and you start to get
wierd problems like the above. If would go with the version that is
crashing and start to remove features from your code. At some point you
will find that it stops crashing, at that point the last thing you
removed will likely be the cause of the problem.

Hopefully you will take a lesson from this painful process. Don't mess
with pointers. Use smart pointers if possible.

thanks

john

[Mark]

ne********@gmail.com wrote:

can you explain it in more detail?
thanks

I think what he was trying to say is that we can't help you unless you
post some code ;-)