Does anyone know what the behaviour of std::string.assign(NULL) is? Assigning NULL (using operator) or constructing from NULL is undefined. Does the same apply for this function?
Add a comment
|
2 Answers
The standard says this in 21.4.6.3 paragraph 12:
Requires: s points to an array of at least traits::length(s) + 1 elements of charT.
In the character traits requirements Table 62 is says:
X::length(p) std::size_t yields: the smallest i such that X::eq(p[i],charT()) is true. linear
That implies that traits::length() will dereference s and, thus, s can't be a pointer to null. You get undefined behavior.
2 Comments
Derf Skren
Could you tell me where you found that quote? I could only find something similar amongst the Boost documentation.
Derf Skren
Disregard, I had to download a PDF... Had thought that the C++ guys would want people to see the standard, but clearly they don't - just Google that phrase!
It's undefined behavior to assign a C++ string from a NULL "C string", because a null pointer is not actually pointing to a C string at all. I could find no reference saying that std::string should check for NULL--you need to do it yourself.
1 Comment
Derf Skren
Thanks. I'm aware that what this code attempts to do is 'morally' wrong, however technically there's no actual reason why memory location 0 can't contain a null-terminated string. This is why I wondered if 'modern' interpretations of this, featuring a properly defined nullptr, might actually catch the error rather than provoke the hardware/OS.