Hide comments about constants in MIR dumps if they bring no additional information

[oli-obk]

cc @lcnr @eddyb

We sometimes have super verbose things in MIR like

         assert(const true, "index out of bounds: the len is {} but the index is {}", const 4_usize, const 2_usize) -> bb1; // scope 0 at $DIR/array_index.rs:5:18: 5:33
                                          // ty::Const
                                          // + ty: bool
                                          // + val: Value(Scalar(0x01))
                                          // mir::Constant
                                          // + span: $DIR/array_index.rs:5:18: 5:33
                                          // + literal: Const { ty: bool, val: Value(Scalar(0x01)) }
                                          // ty::Const
                                          // + ty: usize
                                          // + val: Value(Scalar(0x0000000000000004))
                                          // mir::Constant
                                          // + span: $DIR/array_index.rs:5:18: 5:33
                                          // + literal: Const { ty: usize, val: Value(Scalar(0x0000000000000004)) }
                                          // ty::Const
                                          // + ty: usize
                                          // + val: Value(Scalar(0x0000000000000002))
                                          // mir::Constant
                                          // + span: $DIR/array_index.rs:5:18: 5:33
                                          // + literal: Const { ty: usize, val: Value(Scalar(0x0000000000000002)) }

These constants give no useful additional information, as everything is already contained in the mir terminator. I think we could start out with just not printing these statements if the type is an integer, bool or char, which should cover most use cases. A more complex analysis could later look at constant in more detail and decide whether the verbose printing could be helpful.

[mglagla]

I'd like to try this.

[oli-obk]

ooh, sorry, evanrichter (idk their github handle) is working on it already since yesterday.

[evanrichter]

@mglagla we can work together if you want, as I'm new to rustc dev. would you like to claim by saying @rustbot claim?

[mglagla]

@evanrichter No, it's alright, I merely wanted to test if I could find the relevant part of code to edit and I did. I only would make a pull request if nobody worked on this already.

[evanrichter]

@mglagla sounds like you are ahead of me and I am busy this week. If the issue is still available next week I'll claim it

[xldenis]

@evanrichter do think you'll have time to work on this?

[evanrichter]

@xldenis not until Wednesday at least

[alasher]

Hey guys, sorry to snipe this - there hadn't been any activity in a few weeks so I had a go at fixing this.

@oli-obk I've implemented your suggestion in the linked PR. One note though is that your post above did not specify if we should suppress only ty::Const messages, only mir::Constant messages, or both. My change removes them both for the types you listed, but I could see why we'd want to keep mir::Constant messages. I'm not familiar enough with the project to know what people would prefer, I'm happy to change it to what you think is best.