generated docs improvements: add vector, fntype, anytype type-printing; bool, function value-printing

[s-ol]

See #3404.

I think that some of these aren't currently used in (tested parts of) the stdlib, so this patch can be used to see them in action:

diff --git a/lib/std/meta.zig b/lib/std/meta.zig
index 79b0424e9..719fe1f65 100644
--- a/lib/std/meta.zig
+++ b/lib/std/meta.zig
@@ -702,6 +702,41 @@ pub fn Vector(comptime len: u32, comptime child: type) type {
     });
 }

+pub fn dummy_vec() Vector(3, f32) {
+    var asd : Vector(3, f32) = undefined;
+    return asd;
+}
+
+const OpaqueType = opaque {};
+pub fn dummy_opaque() *OpaqueType {
+    return @intToPtr(*OpaqueType, 1234);
+}
+
+pub fn dummy_fnframe() @Frame(dummy_vec) {
+    return undefined;
+}
+
+pub fn dummy_anyframe() anyframe {
+    return @frame();
+}
+
+pub fn dummy_anyframe_ret() anyframe->f32 {
+    return undefined;
+}
+
+pub fn DummyTypeFnArg(arg: @TypeOf(dummy_opaque)) type {
+    return i8;
+}
+
+test "test vector" {
+    _ = dummy_vec();
+    _ = dummy_opaque();
+    _ = dummy_fnframe();
+    _ = dummy_anyframe();
+    _ = dummy_anyframe_ret();
+    _ = DummyTypeFnArg(dummy_opaque);
+}
+
 /// Given a type and value, cast the value to the type as c would.
 /// This is for translate-c and is not intended for general use.
 pub fn cast(comptime DestType: type, target: anytype) DestType {

[s-ol]

I just found out way too late about getCanonDeclPath, which I essentially reimplemented by exposing parent/container on struct-types and decls respectively. It seems that my new method catches some things getCanonDeclPath doesn't, but I'm not convinced it's any better. I'll have to refactor some of the last commits to merge the two approaches one way or another.

EDIT: dropped my appraoch and switched to using getCanonDeclPath in all cases. If anyone is interested in looking into the other approach, its still visible here: s-ol@a506269

[andrewrk]

Thanks for the patch! The changes look good to me, but the CI failures are legit. Have a look there, looks like some pretty straightforward compile errors.

[s-ol]

@andrewrk thanks for the reminder, I hadn't checked them again after rebasing, and before that it was masked by the LLVM version change breaking CI.

Passing the pointer instead of the value to jw_bool inline 711 was a legit mistake, but I can't figure out where the duplicate case ZigTypeIdBool: is supposed to come from. My code doesn't have it, and I also don't see any such code added to master that could clash in a merge?

[s-ol]

Strange. I can't build locally right now because LLVM hasn't landed on Arch yet, so maybe this has to wait until I can see what's happening for myself.

[LemonBoy]

My code doesn't have it, and I also don't see any such code added to master that could clash in a merge?

#6736 was merged a few days ago and it added the ZigTypeIdBool case.