Wrong implementation for set_cstr for bounded strings

[vladipus]

Indirectly they would use strncpy_s, but that method DOES NOT allow truncating, which seems needed by the semantic of your method.

[vladipus]

this implementation also seems off and risky, since the last argument is actually count that could be used as s1[count]

[vladipus]

Maybe something like that would do?

[vladipus]

Another problem is that strcpy_s guarantees null-termination, but your bound strings may be non-null terminated.

[vladipus]

Hmm, you are using strlen in you bounded string size method. That is strange and contradicts with my statement above. Now I see you have max_size + 1 as an array. But you still have to refactor the usage of strncpy

[vladipus]

You can even end up with an invalid state since currently:

[vladipus]

You could just or maybe rewrite the m_core_strncpy to M_core_strncpy_s and use it everywhere afterwards.

[vladipus]

this method is also implemented incorrectly. the size of the destination buffer can't be removed that simply and the m_core_strncat is also wrong in case of using strncat_s, semantically wrong.

[vladipus]

An interesting thing. At least in debug build. The strncpy_s actually may write data past zero terminator, so it mast be saved manually and restored (for OOR compatibility.

[P-p-H-d]

The issue with strncpy_s and strncat_s should be fixed in master. This is strange however. All the tests already passed with Visual, and they did trigger this behavior of truncation... So I don't know why they succeed...

this method is also implemented incorrectly. the size of the destination buffer can't be removed that simply

Which method?

[P-p-H-d]

This is strange however. All the tests already passed with Visual, and they did trigger this behavior of truncation... So I don't know why they succeed...

It is because there was a constraint violation (as you notice), the default runtime handler is called which terminates the program and returns a negative error code to the caller! This is not handled by the script which assumes a successful termination. (It is fixed in master).

[vladipus]

Which method? bounded string's equal str p

[vladipus]

Are you sure about your re-implementation? I've coped with the issue with _s variants. Seems like it is not that bad of an idea to use the secure methods on the bounded strings, since their size is statically defined.

[P-p-H-d]

Which method? bounded string's equal str p

I see no issue with _equal_str_p of bounded string.

Are you sure about your re-implementation?

As sure as the tests passed

[vladipus]

I see no issue with _equal_str_p of bounded string.

In case you've also reimplemented the cat method yourself, there might be no issue. But there WAS one.

As sure as the tests passed

I've implemented _s variants through non-_s like so and all tests have also passed ;)

[P-p-H-d]

In case you've also reimplemented the cat method yourself, there might be no issue. But there WAS one.

Yes, the service that used the cat method was wrong. But _equal_str_p doesn't use this cat method :)

[vladipus]

Maybe I've misrememeber =) Ok, closing the issue.