Regression when using the volatile module on devel

[Abathargh]

Description

Using the voilatile module currently incurs in a compiler error.

I have run a git bisect and the outcome is the following:

*[72d0ba2df][~/projects/nim/Nim]$ git bisect good
9b378296f65fff962225acc1bef2b3811fffda03 is the first bad commit
commit 9b378296f65fff962225acc1bef2b3811fffda03
Author: ringabout <43030857+ringabout@users.noreply.github.com>
Date:   Wed Apr 10 20:41:16 2024 +0800

    fixes addr/hiddenAddr in strictdefs (#23477)

Minimal example:

import volatile

var ctr = 0

proc main =
  volatileStore(addr ctr, 1)
  echo volatileLoad(addr ctr)

when isMainModule:
  main()

I know this is not a perfect use case for it, but for example I use import volatile to interact with memory mapped registers on mcus, and it's a necessary feature there.

Nim Version

Nim Compiler Version 2.1.1 [Linux: amd64]
Compiled at 2024-04-12
Copyright (c) 2006-2024 by Andreas Rumpf

git hash: 779bc84
active boot switches: -d:release

Current Output

Hint: used config file '/home/mar/projects/nim/Nim/config/nim.cfg' [Conf]
Hint: used config file '/home/mar/projects/nim/Nim/config/config.nims' [Conf]
......................................................................
/home/mar/projects/nim/volatile_test/vol.nim(6, 16) template/generic instantiation of `volatileStore` from here
/home/mar/projects/nim/Nim/lib/pure/volatile.nim(31, 12) Error: type mismatch
Expression: `[]=`(ctr, 1)
  [1] ctr: int
  [2] 1: int literal(1)

Expected one of (first mismatch at [position]):
[0] proc `[]=`(s: var string; i: BackwardsIndex; x: char)
[0] proc `[]=`[I: Ordinal; T, S](a: T; i: I; x: sink S)
[0] proc `[]=`[Idx, T; U, V: Ordinal](a: var array[Idx, T]; x: HSlice[U, V];
                                  b: openArray[T])
[0] proc `[]=`[Idx, T](a: var array[Idx, T]; i: BackwardsIndex; x: T)
[0] proc `[]=`[T, U: Ordinal](s: var string; x: HSlice[T, U]; b: string)
[0] proc `[]=`[T; U, V: Ordinal](s: var seq[T]; x: HSlice[U, V]; b: openArray[T])
[0] proc `[]=`[T](s: var openArray[T]; i: BackwardsIndex; x: T)
[0] template `[]=`(a: WideCStringObj; idx: int; val: Utf16Char)
[0] template `[]=`(s: string; i: int; val: char)

Expected Output

Hint: used config file '/home/mar/.choosenim/toolchains/nim-2.0.2/config/nim.cfg' [Conf]
Hint: used config file '/home/mar/.choosenim/toolchains/nim-2.0.2/config/config.nims' [Conf]
.......................................................................
CC: ../../../.choosenim/toolchains/nim-2.0.2/lib/system/exceptions.nim
CC: ../../../.choosenim/toolchains/nim-2.0.2/lib/std/private/digitsutils.nim
CC: ../../../.choosenim/toolchains/nim-2.0.2/lib/system/dollars.nim
CC: ../../../.choosenim/toolchains/nim-2.0.2/lib/system.nim
CC: vol.nim
Hint:  [Link]
Hint: mm: orc; threads: on; opt: none (DEBUG BUILD, `-d:release` generates faster code)
27658 lines; 0.428s; 30.309MiB peakmem; proj: /home/mar/projects/nim/volatile_test/vol.nim; out: /home/mar/projects/nim/volatile_test/vol [SuccessX]

Possible Solution

No response

Additional Information

No response

[ringabout]

proc semArrPut(c: PContext; n: PNode; flags: TExprFlags): PNode =
  # rewrite `[]=`(a, i, x)  back to ``a[i] = x``.
  let b = newNodeI(nkBracketExpr, n.info)
  b.add(n[1].skipAddr)

I don't think it should skip addr here or it should handle addr deref here and it wasn't skipped before because addr ... existed in the form of a magic function.

[ringabout]

Besides 8990626 changed the meaning of skipAddr for frontend