Support ? mark in nightlies as alternative to try!

[tupshin]

No description provided.

[nrc]

Should be behind an option until ? is stable.

[kamalmarhubi]

This is behind the question_mark lang feature.

[marcusklaas]

This makes method chain formatting a bit more complex. Looking into this.

[kamalmarhubi]

Should be behind an option until ? is stable.

Shouldn't we format code that uses it? Or use the #![feature(question_mark)] as a mark that the code is opting in.

[marcusklaas]

Yea, I don't think it needs to be behind an option either.

[kamalmarhubi]

Having written that, I realize that if we are formatting individual files, we'd also have to find the crate root to see what features are in use. I kind of lean towards just accepting it unconditionally anyway.

[marcusklaas]

Me too.

[nrc]

To clarify I meant that there should be an option which changes try!(expr) to expr?. Just formatting expr? doesn't need to be behind an option,

[mcarton]

Pinging that issue, rustfmt produces weird diffs on Clippy:

@@ -57,11 +57,11 @@ impl<'a, 'tcx, 'v> Hir2Qmm<'a, 'tcx, 'v> {
         for a in a {
             if let ExprBinary(binop, ref lhs, ref rhs) = a.node {
                 if binop.node == op {
-                    v = self.extract(op, &[lhs, rhs], v)?;
+                    v = self.extract(op, &[lhs, rhs], v)?;;
                     continue;
                 }
             }
-            v.push(self.run(a)?);
+            v.push(self.run(a)?));
         }
         Ok(v)
     }
@@ -70,22 +70,22 @@ impl<'a, 'tcx, 'v> Hir2Qmm<'a, 'tcx, 'v> {
         // prevent folding of `cfg!` macros and the like
         if !in_macro(self.cx, e.span) {
             match e.node {
-                ExprUnary(UnNot, ref inner) => return Ok(Bool::Not(box self.run(inner)?)),
+                ExprUnary(UnNot, ref inner) => return Ok(Bool::Not(box self.run(inner)?))),
                 ExprBinary(binop, ref lhs, ref rhs) => {
                     match binop.node {
-                        BiOr => return Ok(Bool::Or(self.extract(BiOr, &[lhs, rhs], Vec::new())?)),
-                        BiAnd => return Ok(Bool::And(self.extract(BiAnd, &[lhs, rhs], Vec::new())?)),
-                        _ => {},
+                        BiOr => return Ok(Bool::Or(self.extract(BiOr, &[lhs, rhs], Vec::new())?))),
+                        BiAnd => return Ok(Bool::And(self.extract(BiAnd, &[lhs, rhs], Vec::new())?))),

[marcusklaas]

@mcarton: is that on the master branch, or using #893?

Edit: either way, it's a bug in the rustc parser that has been fixed, but is still in syntex.

[mcarton]

@marcusklaas didn’t see that PR, just tested with it, works as expected 👍

[marcusklaas]

Closed by #893.