On Tuesday, 24 May 2016 at 23:47:25 UTC, mogu wrote:
I use dub 0.9.25, DMD64 D Compiler v2.071.0. Create a minimal
project and write the following codes in main function.
```
import std.math;
long i = 0;
i += 3.5.floor;
```
Then run dub will get a warning and dmd failed.
```
source/app.d(4,7): Warning: long += double is performing
truncating conversion
dmd failed with exit code 1.
```
This is because dub is apparently compiling with warnings
enabled. If you compile manually with -w, you will get the same
result.
But build the app.d directly use dmd. It's OK. However if the
last line of code above is `i = 3.5.floor;`. An error occurs.
```
Error: cannot implicitly convert expression (floor(3.5)) of
type double to long
```
So what's the deal?
Take a look at [1] and you'll see that floating point types are
not implicitly convertible to integrals, so the error is
appropriate.
As for the += situation, because one operand is a double, the
other will be converted to a double for the operation. Because of
the implicit conversion rules, it may seem that an error should
occur because of the assignment part in the +=. This isn't
something I've really considered before, but I assume the rule is
ignored here because += would not be very useful if you had to
cast the result back to the original type. Where would you place
the cast? That's why you get the truncation warning when
compiling with -w instead of an error about implicit conversions.
https://dlang.org/spec/type.html#usual-arithmetic-conversions
