[fpc-pascal] Does SetLength clear its elements to zeros?

[Mark Morgan Lloyd]

Jonas Maebe wrote:
> Dennis wrote on Fri, 11 Dec 2015:
> 
>> I just tested, SetLength a string  does not zero its elements (the 
>> chars).
>>
>> However, SetLength a dynamic array seems to zero its elements.
> 
> Dynamic arrays are indeed zeroed when changing the length, while strings 
> are not. For dynamic arrays, it is necessary anyway in case it has 
> reference counted elements. Maybe for consistency, it's always done.
> 
> Additionally, changing the length of the string without afterwards 
> overwriting the data between the previous and new last character seems 
> something that would happen very seldom (why did you increase the length 
> if you didn't need to write all characters -- any operation with the 
> string, such as writing or concatenating it, will use those characters 
> anyway because the length indicates they're part of the string).
> 
> The above are mostly guesses though.

Setlength(zero) would be significant if somebody were trying to wipe a 
field that's just been used for password verification.

I'd suggest that this is related to the general issue of whether a 
string that is reallocated due to a substantial length increase is 
wiped, which I think was tentatively fixed by providing an alternative 
memory manager (discussed back when "Heartbleed" was an issue).

I notice that you can now get memory analysis tools that report when a 
block is either more or less random than would be expected, suggesting 
that zeroing sensitive data might not in fact be the best approach.

-- 
Mark Morgan Lloyd
markMLl .AT. telemetry.co .DOT. uk

[Opinions above are the author's, not those of his employers or colleagues]