On Sunday, 8 March 2015 at 15:41:23 UTC, Meta wrote:
template canBeAlias(T...)
if (T.length == 1)
{
static if (is(typeof({alias _ = T[0];})))
{
enum canBeAlias = true;
}
else
{
enum canBeAlias = false;
}
}
pragma(msg, canBeAlias!canBeAlias); //prints "true"
static assert(!canBeAlias!true); //passes
static assert(canBeAlias!(canBeAlias!canBeAlias)); //passes?!
What is going on here? `canBeAlias!canBeAlias` evaluates down
to true, so why is `canBeAlias!true` false when
`canBeAlias!(canBeAlias!canBeAlias)` is true?
I get an error on your code: "test.d(16): Error: static assert
(canBeAlias!(true)) is false". But when commenting out the first
assert (line 15), there's no error.
Played around with it, and I think it's a bug in the compiler:
----
/* The differently numbered 'canBeAliasN' are all the same. */
enum true_ = true;
enum canBeAlias1(T...) = is(typeof({alias _ = T[0];}));
pragma(msg, canBeAlias1!true); /* "false" */
enum canBeAlias2(T...) = is(typeof({alias _ = T[0];}));
pragma(msg, canBeAlias2!true_); /* "true" */
enum canBeAlias3(T...) = is(typeof({alias _ = T[0];}));
pragma(msg, canBeAlias3!true, " ", canBeAlias3!true_); /* "false
false" */
enum canBeAlias4(T...) = is(typeof({alias _ = T[0];}));
pragma(msg, canBeAlias4!true_, " ", canBeAlias4!true); /* "true
true" */
----
On their own, `canBeAlias!true` = false and `canBeAlias!true_` =
true. This makes sense, because `true_` is a symbol whereas
`true` is a value.
But the two instantiations are apparently recognized as
equivalent. Whichever is instantiated first, its value is used
for the other instantiation, too. I think this behaviour is wrong.
