Comment #2 by bearophile_hugs — 2011-08-22T23:21:42Z
The given code example now works. But this:
T foo(T)() {
T[1] a = [1];
a[] += a[];
return a[0];
}
static assert(foo!long() == 2); // OK
static assert(foo!int() == 2); // error
void main() {}
Gives:
test.d(3): Error: _arraySliceSliceAddass_i cannot be interpreted at compile time, because it has no available source code
test.d(7): Error: cannot evaluate foo() at compile time
test.d(7): Error: static assert (foo() == 2) is not evaluatable at compile time
This behaviour difference between long and int is not good. Do you prefer me to reopen this bug or to open a new one?
Comment #3 by clugdbug — 2011-08-23T01:29:13Z
(In reply to comment #2)
> The given code example now works. But this:
>
>
> T foo(T)() {
> T[1] a = [1];
> a[] += a[];
> return a[0];
> }
> static assert(foo!long() == 2); // OK
> static assert(foo!int() == 2); // error
> void main() {}
>
>
> Gives:
>
> test.d(3): Error: _arraySliceSliceAddass_i cannot be interpreted at compile
> time, because it has no available source code
> test.d(7): Error: cannot evaluate foo() at compile time
> test.d(7): Error: static assert (foo() == 2) is not evaluatable at compile
> time
>
> This behaviour difference between long and int is not good. Do you prefer me to
> reopen this bug or to open a new one?
A new one. Surprisingly, it actually has *nothing* in common with this bug.
This bug was actually a problem with array slicing, and didn't require array operations.
The new case is that the array operations which are special-cased in the runtime don't have special case treatment in the CTFE engine, since CTFE doesn't have access to the source code of druntime. The array ops which aren't special-cased don't use druntime code, so they should all work.
Comment #4 by bearophile_hugs — 2011-08-23T02:10:18Z
(In reply to comment #3)
> A new one.
OK. It is bug 6545