Syntax errors in julia nightly

[longemen3000]

hello,

in our routine testset run, suddently all nightly tests stopped working with a loading error:

Failed to precompile ClapeyronMultiComponentFlashExt [b5da60bd-4eec-5f7e-9c86-09fe8b672af4] to "/home/runner/.julia/compiled/v1.12/ClapeyronMultiComponentFlashExt/jl_Ey1fLn".
ERROR: LoadError: syntax: "(Expr(:escape, :var"##collection#274"),)" is not a valid function argument name around /home/runner/.julia/packages/Roots/E1WQf/src/Derivative/lith.jl:268

maybe it is related to #57188 ?

[Seelengrab]

Hi! Do you have a minimal example showing the error?

[longemen3000]

It appears when using Roots.jl, give me a moment to craft the MWE

[longemen3000]

smaller MVE (just with Accessors.jl)

using Pkg
Pkg.add("Accessors")
using Accessors
struct TestX{N,T}
  a::NTuple{N,T}
end
A = TestX((1,1,2))
@reset A.a[end] = 4 #this line causes the error

[macd]

I think this is due to the broken indexing into a tuple. Ie:

julia 1.12> a = (1,2,3)
(1, 2, 3)

julia 1.12> a[1]┌ Error: Error in the keymap
│   exception =
│    UndefVarError: `check_for_hint` not defined in `REPL.LineEdit`
│    Suggestion: check for spelling errors or missing imports.
│    Stacktrace:
│     [1] getproperty
│       @ ./Base_compiler.jl:47 [inlined]
...

[longemen3000]

It is? Seems like a syntax error is fundamentally a different error than the hint error shown above

[macd]

I think the hint error is a red herring, but I could be wrong. This error occurs in Setfield which propagates to Polynomials etc, which is what I thought was happening here.

…

On Sat, Feb 1, 2025 at 6:03 PM Andrés Riedemann ***@***.***> wrote: It is? Seems like a syntax error is fundamentally a different error than the hint error shown above — Reply to this email directly, view it on GitHub <#57223 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AADKDRVRNVP5N7MCVMHMFUL2NV4F7AVCNFSM6AAAAABWH3PCOCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMMRZGIYDENJSGM> . You are receiving this because you commented.Message ID: ***@***.***>

[inkydragon]

Introduced by #57188, See #57188 (comment)
Latest PkgEval shows that this change broken many pkgs: https://s3.amazonaws.com/julialang-reports/nanosoldier/pkgeval/by_date/2025-02/02/report.html

MWE:

macro t57223()
    :( $(esc(:x))->$(esc(:x)) )
end
@t57223()(111)
# ERROR: syntax: "(Expr(:escape, :x),)" is not a valid function argument name around REPL[2]:1

Or

macro t57223()
    :( $(:x)->0 )
end
@t57223()(111)
# ERROR: syntax: "(:x,)" is not a valid function argument name around REPL[9]:1

So seems something wrong with escape in macro $().

# Julia Version 1.12.0-DEV.1970
julia> :( $(:x)->0 )
:((:x,)->begin
          #= REPL[21]:1 =#
          0
      end)

julia> Meta.parse(":( $(:x)->0 )")
:($(Expr(:quote, :(x->begin
          #= none:1 =#
          0
      end))))

julia> Meta.lower(Main, Meta.parse(":( $(:x)->0 )"))
:($(Expr(:thunk, CodeInfo(
1 ─ %1 = $(Expr(:copyast, :($(QuoteNode(:(x->begin
          #= none:1 =#
          0
      end))))))
└──      return %1
))))

julia> macro t57223()
           :( $(:x)->0 )
       end
@t57223 (macro with 8 methods)

julia> @macroexpand1 @t57223()(111)
:(((((:x,),)->begin
          #= REPL[25]:2 =#
          0
      end))(111))

Note that the arguments to the anonymous function become tuples.

# Julia Version 1.11.3
julia> :( $(:x)->0 )
:(x->begin
          #= REPL[9]:1 =#
          0
      end)

julia> Meta.parse(":( $(:x)->0 )")
:($(Expr(:quote, :(x->begin
          #= none:1 =#
          0
      end))))

julia> Meta.lower(Main, Meta.parse(":( $(:x)->0 )"))
:($(Expr(:thunk, CodeInfo(
    @ none within `top-level scope`
1 ─ %1 = $(Expr(:copyast, :($(QuoteNode(:(x->begin
          #= none:1 =#
          0
      end))))))
└──      return %1
))))

julia> macro t57223()
           :( $(:x)->0 )
           end
@t57223 (macro with 1 method)

julia> @macroexpand1 @t57223()(111)
:((((var"#2#x",)->begin
          #= REPL[11]:2 =#
          0
      end))(111))

It seems that both the parsing and lowering phases generate the same output.
I wonder if this error might have nothing to do with JuliaSyntax.
Maybe there is something wrong with Macro hygiene or macroexpand.

[KristofferC]

That PkgEval looked quite bleak indeed, I think we have to revert for now?