https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85861
--- Comment #13 from Jonny Grant <jg at jguk dot org> ---
(In reply to Jonathan Wakely from comment #11)
> My guess is that we don't want to warn about conversions that are
> well-defined and the original value can be obtained by a round-trip.
> Converting a size_t to an int is lossy, i.e. converting back to size_t may
> not give the original value. But converting an int to size_t and back to an
> int is value preserving.
Do you mean because size_t is 64bit typically, and int is only 32bit?
I tested this on my 64bit Ubuntu
It converts a size_t 9,223,372,036,854,775,807 to an int
and ends up with 18,446,744,073,709,551,615
Which I was surprised about, as I thought int was 4 bytes. I had expected
4,294,967,295
printf("size_t: %zu\n", sizeof(size_t));
printf("int: %zu\n", sizeof(int));
size_t big = 9223372036854775807;
int smaller = big;
size_t big2 = smaller;
printf("smaller: %d, big2: %zu\n", smaller, big2);
> So I can see why it makes sense to have different warnings for lossy and
> non-lossy conversions.
That sounds good, all lossy would be useful to have a loss of precision
warning.
