[PATCH] D97311: [DebugInfo] Introduce DIConstant metadata for representing named constants

[Sourabh Singh Tomar via Phabricator via llvm-commits]

SouraVX added a comment.

In D97311#2588856 <https://reviews.llvm.org/D97311#2588856>, @aprantl wrote:

> Have you considered instead introducing a `bool isConstant` field in DILocalVariable? This way we could reuse all the machinery for local/global variables and only need to switch out the TAG in AsmPrinter.

Thanks for the review, @aprantl!
I considered this approach while hacking/iterating, but as mentioned here:
https://lists.llvm.org/pipermail/llvm-dev/2021-February/148447.html

  A naïve attempt to represent it as GlobalVariable(or constant) as
  
  !7 = !DIGlobalVariableExpression(var: !8, expr: !DIExpression(DW_OP_consts, 200))
  !8 = distinct !DIGlobalVariable(name: "bar", scope: !2, file: !3, line: 3, type: !9, isLocal: false, isDefinition: true)
  
  This will materialize in DWARF as:
  0x0000004a:     DW_TAG_variable
                    DW_AT_name    ("bar")
                    ...
                   DW_AT_location        (DW_OP_addr 0x2007d4, DW_OP_consts +200)  // This is incorrect, pointing to data section

For `DW_OP_consts` case, it was creating an `location` entry like `GlobalVariable`, this wasn't making sense to me(at least), hence I thought of going this route.

Another motivating reason was: (well not an issue rather) introducing `isConstant` field is a bit non-trivial, considering tightly coupled nature of `DILocalVariable` and `DIGlobalVariable` with `DIVariable`. In my first attempt I introduced this `isConstant` field in `DIVariable` so that it can be utilized well for both situation. Obviously it worked - but broke some round tripping test cases.

I'll give it one more shot, if you have any ideas/suggestion WRT approach, please share I'll be happy to reason/iterate/implement.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D97311/new/

https://reviews.llvm.org/D97311