minimal evaluation for loading.jl makes Julia internals broken

[aviatesk]

It seems like the minimal evaluation in Revise isn't working correctly against loading.jl (probably due to an issue with LoweredCodeUtils?).

I've confirmed that this problem exists in v1.12 with the master branch of Revise/LoweredCodeUtils.jl.
I've also seen the issue even when using the PR that updates LCU.jl to be compatible with 1.12.

We can reproduce the problem like this:
If we run maybe_extract_sigs! on loading.jl, the Pkg prompt gets messed up. After that, commands like st will result in errors.

julia> using Revise

julia> let pkgid = Revise.get_tracked_id(Base)
           filename = "loading.jl"
           pkgdata = Revise.pkgdatas[pkgid]
           Revise.maybe_extract_sigs!(Revise.maybe_parse_from_cache!(pkgdata, filename))
       end;
[ Info: tracking Base

pkg> st # weird things happen when entering pkg mode
ERROR: SystemError: getattrlist: Invalid argument
Stacktrace:
  [1] systemerror(p::Symbol, errno::Int32; extrainfo::Nothing)
    @ Base ./error.jl:186
  [2] systemerror
    @ ./error.jl:185 [inlined]
  [3] isfile_casesensitive(path::String)
    @ Base ~/julia/julia4/base/loading.jl:56
  [4] active_project(search_load_path::Bool)
    @ Base ./initdefs.jl:347
  [5] active_project
    @ ./initdefs.jl:340 [inlined]
  [6] find_project_file(env::Nothing)
    @ Pkg.Types ~/julia/julia4/usr/share/julia/stdlib/v1.12/Pkg/src/Types.jl:203
  [7] Pkg.Types.EnvCache(env::Nothing)
    @ Pkg.Types ~/julia/julia4/usr/share/julia/stdlib/v1.12/Pkg/src/Types.jl:382
  [8] EnvCache
    @ ~/julia/julia4/usr/share/julia/stdlib/v1.12/Pkg/src/Types.jl:382 [inlined]
  [9] Context
    @ ~/julia/julia4/usr/share/julia/stdlib/v1.12/Pkg/src/Types.jl:444 [inlined]
 [10] status(pkgs::Vector{Pkg.Types.PackageSpec}; io::IOContext{IO}, kwargs::@Kwargs{})
    @ Pkg.API ~/julia/julia4/usr/share/julia/stdlib/v1.12/Pkg/src/API.jl:155
 [11] status(pkgs::Vector{Pkg.Types.PackageSpec})
    @ Pkg.API ~/julia/julia4/usr/share/julia/stdlib/v1.12/Pkg/src/API.jl:153
 [12] do_cmd(command::Pkg.REPLMode.Command, io::Base.TTY)
    @ Pkg.REPLMode ~/julia/julia4/usr/share/julia/stdlib/v1.12/Pkg/src/REPLMode/REPLMode.jl:409
 [13] do_cmds(commands::Vector{Pkg.REPLMode.Command}, io::Base.TTY)
    @ Pkg.REPLMode ~/julia/julia4/usr/share/julia/stdlib/v1.12/Pkg/src/REPLMode/REPLMode.jl:395
 [14] do_cmds(repl::REPL.LineEditREPL, commands::String)
    @ REPLExt ~/julia/julia4/usr/share/julia/stdlib/v1.12/Pkg/ext/REPLExt/REPLExt.jl:99
 [15] on_done(s::REPL.LineEdit.MIState, buf::IOBuffer, ok::Bool, repl::REPL.LineEditREPL)
    @ REPLExt ~/julia/julia4/usr/share/julia/stdlib/v1.12/Pkg/ext/REPLExt/REPLExt.jl:113
 [16] apply_gr(::GlobalRef, ::Any, ::Vararg{Any})
    @ Base ./reflection.jl:1278
 [17] #invokelatest_gr#232
    @ ./reflection.jl:1285 [inlined]
 [18] invokelatest_gr
    @ ./reflection.jl:1281 [inlined]
 [19] (::REPLExt.var"#create_mode##0#create_mode##1"{REPL.LineEditREPL})(s::REPL.LineEdit.MIState, buf::IOBuffer, ok::Bool)
    @ REPLExt ~/julia/julia4/usr/share/julia/stdlib/v1.12/Pkg/ext/REPLExt/REPLExt.jl:135
 [20] run_interface(terminal::REPL.Terminals.TextTerminal, m::REPL.LineEdit.ModalInterface, s::REPL.LineEdit.MIState)
    @ REPL.LineEdit ~/julia/julia4/usr/share/julia/stdlib/v1.12/REPL/src/LineEdit.jl:2852
 [21] run_frontend(repl::REPL.LineEditREPL, backend::REPL.REPLBackendRef)
    @ REPL ~/julia/julia4/usr/share/julia/stdlib/v1.12/REPL/src/REPL.jl:1657
 [22] (::REPL.var"#61#62"{REPL.LineEditREPL, REPL.REPLBackendRef})()
    @ REPL ~/julia/julia4/usr/share/julia/stdlib/v1.12/REPL/src/REPL.jl:648

[timholy]

Have you checked this on #894 yet? I've seem some things like this before but not since I've been running with that PR. But I don't think I checked loading.jl specifically.

[aviatesk]

Whoops, this issue seems to have been fixed on #894!

[timholy]

No "whoops" needed, that's not merged yet.

My guess is that it's the change in #894 that made

Revise.jl/src/lowered.jl

Lines 454 to 456 in 5105f0a

	elseif LoweredCodeUtils.get_lhs_rhs(stmt) !== nothing
	# If we're here, either isrequired[pc] is true, or the mode forces us to eval assignments
	pc = step_expr!(recurse, frame, stmt, true)

conditional on mode:

               if mode === :sigs && stmt.head === :const && (a = stmt.args[1]) isa GlobalRef && @invokelatest(isdefined(mod, a.name))
                    # avoid redefining types unless we have to
                    pc = next_or_nothing!(frame)
                else
                    pc = step_expr!(recurse, frame, stmt, true)
                end

I guess that wasn't an issue before bpart, but we could consider making that change independently of #894.

[aviatesk]

Yes, that's right.
I confirmed that changing that part fixes the issue in master with the above repro.
It would be very helpful if you could merge this soon, separate from #894.

[timholy]

xref #910

[timholy]

JuliaRegistries/General#129431