etl::string is invalid if calling assign() with itself

[jbenczarski]

If you assign a string to itself, the first character of the string is set to '0' because initialise() is called during assign(). Is this intentional?

Simple self assignment:

etl::string<32> str1 = "hello world";
str1.assign(str1);

Assignment to itself via string_view of itself:

etl::string<32> str = "hello world";
auto sv = etl::string_view(str);
str.assign(sv.begin(), sv.end());

[jwellbelove]

The first could be solved with a this != &other test.

The second is more problematic. The only thing you could efficiently check is a data() != view.data() test, but that would fail where view.begin() isn't at the start of the string.

[jbenczarski]

Well, the issue only exists when you assign starting from (str/view).begin().

While not very efficient, would it work if assign did not call initialise() and instead only set current_size = 0? I believe it would fix my issue, but I'm not sure of all the ramifications.

[jwellbelove]

I think that could work. It would have to erase the unused portion of the string if the 'secure' flag was set.

[jwellbelove]

After some thought and experimentation, it is possible to stop a self assign from destroying the string, but it isn't possible to get etl::string_view parameter to work in all circumstances (reverse iterators).

The only option would be to throw an error if the address of the first element is within the address range of the string's buffer. (This has it's own issues)

The STL doesn't do this, the standard just says that elements must not be within the address range of the string's buffer.

Caveat Emptor.

[jwellbelove]

Fixed 20.40.0