On Wednesday, 29 April 2015 at 17:35:58 UTC, Andrei Alexandrescu
wrote:
I have no doubt the change can find certain errors. Problem is
false positives. FWIW these are the changes I had to operate on
std.allocator to make it work with the new compiler. One per
194 lines on average, all false positives:
Yeah, but I think that it's safe to say that std.allocator is not
the normal case, since it's operating so heavily on arrays (and
non-GC allocated no less), and it's caring about their pointers
rather than what they point to, whereas most D array code doesn't
care about the ptr value of the array at all. In general, at
most, it'll care about whether the array is empty or not. So, I
would expect that for most programs, they won't see this warning
much - be it a false positive or not - and if they do see it, the
odds are pretty high that it's not a false positive. Yes, there
will be code out there which uses if(arr) correctly, and not
being able to do if(auto ptr = arr) will be annoying for some
folks, but given how null and empty are so frequently conflated
with arrays in D, I very much doubt that much code is going to be
using if(arr) and want it to be equivalent to if(arr !is null).
Granted, we can't know for sure how common if(arr) is used in the
wild - be it correctly or incorrectly - but std.allocator is
definitely not your average code, and the consensus seems to be
that if(arr) is error-prone even if it's occasionally useful.
- Jonathan M Davis
