Thread: string operation and related exception

Hi Daved,


In release mode, this line will check the boundary. Any comments?

Code:

_SCL_SECURE_VALIDATE_RANGE(_Pos < size());

Originally Posted by Daved

The release version only has range checking if the _HAS_ITERATOR_DEBUGGING symbol is defined and set to 1. My guess is that by default it is not, so by default the release version skips the range check.

So what you're seeing is that the code follows what CornedBee stated, and additionally implemented Elysia's preference that there be a switch of some kind to allow boundary checking even in release build.

regards,
George

Can you post the definition of _SCL_SECURE_VALIDATE_RANGE?

Sure, Daved.


Here it is. Any comments?

Code:

 #define _SCL_SECURE_VALIDATE_RANGE(cond)				\
	{													\
		if (!(cond))									\
		{												\
			_ASSERTE((#cond, 0));						\
			_SCL_SECURE_OUT_OF_RANGE_NO_ASSERT;			\
		}												\
		__analysis_assume(cond);						\
	}

Originally Posted by Daved

Can you post the definition of _SCL_SECURE_VALIDATE_RANGE?

regards,
George

>> Here it is. Any comments?
Question: Is it always defined like that, or is its definition dependent on some other symbol in the way that _HAS_ITERATOR_DEBUGGING is?

I think always, Daved. If I am not correct, please feel free to correct me. Any comments?

Originally Posted by Daved

>> Here it is. Any comments?
Question: Is it always defined like that, or is its definition dependent on some other symbol in the way that _HAS_ITERATOR_DEBUGGING is?

regards,
George

Hi Elysia,


Three more comments,

1. What is your conclusion? operator[] can throw in release mode?

2. I do not know quite understand what do you mean in the following sentence,

--------------------
In Debug, operator [] throws a fit several times (ASSERTs), does not throw.
--------------------

3. How _SECURE_SCL is defined or not? Is it implicitly defined when some other commonly used macros are defined?

You mentioned both throws and not throw. Sorry English is not my native language, what do yo mean? :-)

Originally Posted by Elysia

On closer inspection, if _SECURE_SCL is defined to 1, _SCL_SECURE_VALIDATE_RANGE is defined as George2 has posted. Otherwise it is simply defined as _SCL_SECURE_VALIDATE_RANGE.
Nothing, in other words.
So far as I see, _SECURE_SCL is not dependant on Release/Debug and defaults to 1 if not defined. It may be defined in other headers, however. Not sure.

In Debug, operator [] throws a fit several times (ASSERTs), does not throw.
at() throws in both debug and release.
if _SECURE_SCL is not defined, then release ignores invalid subscript ranges (go figure) (release).
Debug still complains about out of range.
And lastly, iterator debugging is defined in both release/debug and _SECURE_SCL is always defined to one inside vector header.

regards,
George

It still seems as if you can govern whether bounds checking is done with the _SECURE_SCL symbol.

Although given Microsoft's recent attention to security and avoiding things like buffer overruns, it wouldn't be that surprising if it now defaulted to range checking in release build.

Have you looked at the Microsoft documentation? That might have better information than trying to read though the code itself.

Originally Posted by Elysia

I think either I would allow a flag to be set (EnableExceptions) or differentiate two types (perhaps just different names for initializing a constructor in different ways.

A flag sucks because somebody has to set it, and you never know which way it was set by the person before you.

vector(bool bEnableExceptions = false)

So now I have to specify a fundamental design decision (whether the operator[] does bounds checks) every time I declare a vector?

#define SafeVector vector(true)

I literally barfed on that one..

I would expect the operator [] and the member function at both to throw exceptions if exceptions are enabled. Otherwise if they aren't enabled, none should throw (or make out of bounds checks).

I expect cars to have seven cylinders, and be built entirely of fiberglass.

I understand they want to keep C compability, but then you would be using a C array and not std::vector since that wouldn't even compile under C so not C project would ever use std::vector.

Why would I want to give up all the conveniences of a std::vector and go back to dumb arrays just because I don't want bounds checking?

Thanks Elysia,


1.

Originally Posted by Elysia

Yep, there's a whole topic on the subject. I have to study it and divulged all information first before reporting back.

Could you provide an URL please?

2.

I am wondering how assert is implemented internally? Using some soft interrupt or through by exception handling approach?


regards,
George

Thanks Elysia,

Originally Posted by Elysia

They use a hardware interrupt, __asm int 3 on Intel & AMD processors.

I know int 3 will make a H/W interrupt. And then invoke the interruption handler for int 3. I am wondering how does the assert dialog opens? Visual Studio registers or Windows registers the interrupt handler, and the interrupt handler's task is to display the dialog and some debug information?


regards,
George