methoddef!: use mt => sig format when filling in signatures

[serenity4]

This allows Revise to work with timholy/CodeTracking.jl#140. Requires JuliaDebug/JuliaInterpreter.jl#680.

This change is breaking, so we may either:

• Tag a new breaking release.
• Keep backwards compatibility by supporting a signatures vector that uses the previous sig format.

[timholy]

Looks great. You'd want to bump the [compat] for CodeTracking to 2.

I think I'd favor the breaking release, but if you see advantages in keeping backwards compatibility then I wouldn't stand in the way.

[timholy]

Maybe just use the one in CodeTracking?

[serenity4]

Looks great. You'd want to bump the [compat] for CodeTracking to 2.

CodeTracking is not a direct dependency of LoweredCodeUtils, but since it indirectly is via JuliaInterpreter I'll add it.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 77.09%. Comparing base (550a2a4) to head (b8520fc).
Report is 35 commits behind head on master.

Additional details and impacted files

@@             Coverage Diff             @@
##           master     #125       +/-   ##
===========================================
- Coverage   88.69%   77.09%   -11.61%     
===========================================
  Files           6        6               
  Lines        1442     1528       +86     
===========================================
- Hits         1279     1178      -101     
- Misses        163      350      +187     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[serenity4]

The failing test at https://github.com/JuliaDebug/LoweredCodeUtils.jl/actions/runs/14648771143/job/41109358572?pr=125#step:7:112 asserts that a function object depends on its first method definition. I slightly restructured the implementation for method code edge dependencies in 8548679, and naturally expected a function binding to depend on its declaration, but not on the method definition. @aviatesk or @timholy would that be more correct according to you, or is it a regression?

In the code below, the defintion of Main.ModEval.revise538 now depends on $(Expr(:method, :(Main.ModEval.revise538))) only, and not on the actual method definition:

CodeInfo(
    @ /home/serenity4/.julia/dev/LoweredCodeUtils/test/codeedges.jl:339 within `unknown scope`
1 ─ %1  = enter #4
    @ /home/serenity4/.julia/dev/LoweredCodeUtils/test/codeedges.jl:340 within `unknown scope`
2 ─       global revise538
│         $(Expr(:latestworld))
│         $(Expr(:method, :(Main.ModEval.revise538)))
│         $(Expr(:latestworld))
│         $(Expr(:latestworld))
│   %7  = Main.ModEval.revise538
│   %8  =   dynamic Core.Typeof(%7)
│   %9  = Main.ModEval.Float32
│   %10 =   builtin Core.svec(%8, %9)
│   %11 =   builtin Core.svec()
│   %12 =   builtin Core.svec(%10, %11, $(QuoteNode(:(#= /home/serenity4/.julia/dev/LoweredCodeUtils/test/codeedges.jl:340 =#))))
│         $(Expr(:method, :(Main.ModEval.revise538), :(%12), CodeInfo(
    @ /home/serenity4/.julia/dev/LoweredCodeUtils/test/codeedges.jl:341 within `unknown scope`
1 ─ %1 = Main.ModEval.println
│   %2 =   dynamic (%1)("F32")
└──      return %2
)))
│         $(Expr(:latestworld))
│   %15 = Main.ModEval.revise538
└──       $(Expr(:leave, :(%1)))
3 ─       return %15
4 ┄       e = $(Expr(:the_exception))
│   @ /home/serenity4/.julia/dev/LoweredCodeUtils/test/codeedges.jl:344 within `unknown scope`
│   %19 = Main.ModEval.println
│   %20 =   dynamic (%19)("caught error")
│         $(Expr(:pop_exception, :(%1)))
└──       return %20
)

julia> lr[13]
false

Nothing in the refactor depends on that so I'm fine removing this change if necessary.