Make verify trim printing more readable (also fix the colours) by gbaraldi · Pull Request #60350 · JuliaLang/julia

gbaraldi · 2025-12-09T19:54:14Z

pr:

master:

Inspired by @jakobnissen's blog https://viralinstruction.com/posts/aoc2025/

Written with AI help

topolarity

Looks like a solid improvement to me. Thanks @gbaraldi !

topolarity · 2025-12-11T17:27:55Z

Compiler/src/verifytrim.jl

-        if newstmt.head ∈ (:quote, :inert)
-            return newstmt
+# Check if a union is small with concrete types (same logic as code_warntype's is_expected_union)
+function is_expected_union(@nospecialize(u::Union))


Maybe instead of duplicating, we should move these into the Compiler printing utilities and have InteractiveUtils pull them in from there?

topolarity · 2025-12-11T17:38:02Z

Compiler/src/verifytrim.jl

+end
+
+# Print with light_black if stable, normal otherwise
+function print_styled(io::IO, @nospecialize(x), stable::Bool)


Suggested change

function print_styled(io::IO, @nospecialize(x), stable::Bool)

function print_nontype(io::IO, @nospecialize(x), stable::Bool)

Not a great name, but better than print_styled and printstyled floating around in the same function

Compiler/src/verifytrim.jl

topolarity · 2025-12-11T17:44:46Z

Compiler/src/verifytrim.jl

+# Convert Core.Argument to slot name symbol if available
+function argument_name(codeinfo::CodeInfo, arg::Core.Argument)
+    slotnames = codeinfo.slotnames
+    if slotnames !== nothing && arg.n <= length(slotnames)


The IR verifier seems to suggest that codeinfo.slotnames is always populated and the right length - Is that not true?

Where do you see that?

Most uses I see do check it.

topolarity · 2025-12-11T17:48:01Z

Compiler/src/verifytrim.jl

-    if codeinfo.slotnames !== nothing
-        io = IOContext(io, :SOURCE_SLOTNAMES => sourceinfo_slotnames(codeinfo))
+# Colors for matching parentheses at different nesting levels
+const PAREN_COLORS = (:blue, :yellow, :magenta, :cyan, :green)


I'm really not a fan of multi-colored parentheses personally

They really overload the maximum number of visual items I can track at a time. Do you think they're very helpful?

topolarity · 2025-12-11T17:56:05Z

Compiler/src/verifytrim.jl

+        if warn
+            printstyled(io, "Verifier warning #", no, ": "; color=Base.warn_color(), bold=true)
+        else
+            printstyled(io, "Verifier error #", no, ": "; color=Base.error_color(), bold=true)
+        end


Suggested change

if warn

printstyled(io, "Verifier warning #", no, ": "; color=Base.warn_color(), bold=true)

else

printstyled(io, "Verifier error #", no, ": "; color=Base.error_color(), bold=true)

end

printstyled(io, "Problem #", no, ": "; color=Base.warn_color(), bold=true)

The index is shared but these are counted separately, so changing between "error" / "warning" seems a bit confusing to me

I went with just error here. I don't think the user deciding to ignore errors means we should print them as warnings

topolarity

Looks like you're missing a repr when printing constant values:

Error #1: unresolved call from statement Base.print(
  Base.stdout::IO,
  x::String,

::String # this is "\n"::String
)::Nothing

ran into this with:

function @main(ARGS)
    println("hello!")
end

This can also end up with some pretty complex output at times:

Main.SimpleLib.unsafe_wrap(
  Main.SimpleLib.Array::Type{Array},
  Base.getproperty(
    Base.getproperty(
      Main.SimpleLib.tree::Any,
      …::Symbol
    )::Any,
    :data::Symbol
  )::Any,
  Base.getproperty(
    Base.getproperty(
      Main.SimpleLib.tree::Any,
      …::Symbol
    )::Any,
    :length::Symbol
  )::Any
)::Array{T, N} where {N, T}

I'm not sure unwrapping that many calls is needed? It seems that beyond 1 or 2 you are likely to want to dive into the IR instead.

Oh and the usage of "…" reminds me of Varargs, but that's not what it means. It would be good to make that clearer somehow.

topolarity · 2025-12-16T23:21:49Z

Compiler/src/verifytrim.jl

+function argument_name(codeinfo::CodeInfo, arg::Core.Argument)
+    slotnames = codeinfo.slotnames
+    if slotnames !== nothing && arg.n <= length(slotnames)
+    return slotnames[arg.n]


odd indentation

topolarity

This is looking pretty good now!

My only remaining wishlist is to translate intrinsics back to user-familiar syntax, such as:
- getproperty(x::T, :field)::FieldT -> (x::T).field::FieldT
- Base.memoryrefget((…)::MemoryRef{T}, (…)::Symbol, (…)::Bool) -> (...)[]::T
- etc.

I'm sure there will be plenty of details to work out w.r.t. when we can do that faithfully without mis-leading the user and exactly how we want to translate it all, so probably best left to a separate PR anyway.

Very minor, but I also think it'd be nice to have an extra newline before the stacktrace, like:

Error #1: unresolved call from statement Main.SimpleLib.println(
  Main.SimpleLib.stderr::IO,
  "bad time"::String,
  Base.memoryrefget((…)::MemoryRef{Main.SimpleLib.CTree{Float64}}, (…)::Symbol, (…)::Bool)::Main.SimpleLib.CTree{Float64}
)::Any

Stacktrace:
 [1] size(tree::Main.SimpleLib.CTree{Float64})
   @ Main.SimpleLib ~/repos/julia/test/trimming/libsimple.jl:9
 [2] size(tree::Main.SimpleLib.CTree{Float64})
   @ Main.SimpleLib ~/repos/julia/test/trimming/libsimple.jl:11

Other than that, this looks ready to go to me. Thanks again @gbaraldi !

topolarity · 2025-12-19T15:13:24Z

Compiler/src/verifytrim.jl

+# Use repr for strings/chars to show quotes
 function print_nontype(io::IO, @nospecialize(x), stable::Bool)
-    stable ? printstyled(io, x; color=:light_black) : print(io, x)
+    txt = x isa Union{AbstractString, AbstractChar} ? Base.repr(x) : x


Any reason not to always use repr?

topolarity · 2025-12-19T16:03:28Z

Actually hold on.. This seems to still have pretty aggressive nesting somehow:

Base.prod(
  Base.getfield(
    Main.SimpleLib.unsafe_wrap(
      (…)::Type{Array},
      Base.getproperty(
        Base.getproperty(
          Main.SimpleLib.tree::Any,
          (…)::Symbol
        )::Any,
        (…)::Symbol
      )::Any,
      Base.getproperty(
        Base.getproperty(
          Main.SimpleLib.tree::Any,
          (…)::Symbol
        )::Any,
        (…)::Symbol
      )::Any
    )::Array{T, N} where {N, T},
    (…)::Symbol
  )::NTuple{N, Int64} where N
)::Int64

Is the limit not being applied?

This would almost be readable if we translated the intrinsics back to syntax, but as-is it's pretty overwhelming

topolarity · 2025-12-19T16:05:44Z

Also the (…)::Symbol is hiding some pretty crucial information.. That's a really nice identifier to have.

Can we avoid abbreviating "small" constant values in that position?

gbaraldi · 2026-01-19T16:31:34Z

We may want to port this to 1.13. and maybe even 1.12 (it's technically a bug fix of #60412

topolarity self-requested a review December 9, 2025 23:41

topolarity reviewed Dec 11, 2025

View reviewed changes

Make verify trim printing more readable (also fix the colours)

eca85d5

gbaraldi force-pushed the gb/fancy-printing branch from 9c34e95 to eca85d5 Compare December 11, 2025 21:41

gbaraldi added 3 commits December 11, 2025 22:14

Fix tests

b995968

Fix tests and add coverage

2a66ff6

Fix test again

7e46d60

gbaraldi mentioned this pull request Dec 16, 2025

JuliaC prints verifier errors to stdout, not stderr JuliaLang/JuliaC.jl#85

Closed

Change problem to error

2f9dc68

topolarity requested changes Dec 18, 2025

View reviewed changes

Fix recursion issue and reduce nesting

42a1cbe

topolarity linked an issue Dec 19, 2025 that may be closed by this pull request

Severe memory consumption (20 GB+) when printing certain --trim errors #60412

Closed

Fix indentation

a56756b

topolarity approved these changes Dec 19, 2025

View reviewed changes

gbaraldi added 7 commits December 22, 2025 18:10

Refactor implementation so it's more readable

52add34

Remove dead parens

9ec5dde

Add newline

077dc8d

Use repr for most cases

652c50d

Fix test

5074283

Merge branch 'master' into gb/fancy-printing

6b66d13

Merge branch 'master' into gb/fancy-printing

c858146

gbaraldi added the merge me PR is reviewed. Merge when all tests are passing label Jan 19, 2026

DilumAluthge merged commit 2d9a3f8 into master Jan 20, 2026
9 checks passed

DilumAluthge deleted the gb/fancy-printing branch January 20, 2026 00:14

DilumAluthge removed the merge me PR is reviewed. Merge when all tests are passing label Jan 20, 2026

	function print_styled(io::IO, @nospecialize(x), stable::Bool)
	function print_nontype(io::IO, @nospecialize(x), stable::Bool)

Uh oh!

Conversation

gbaraldi commented Dec 9, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

topolarity left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

gbaraldi Dec 11, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

topolarity left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

topolarity left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

topolarity commented Dec 19, 2025

Uh oh!

topolarity commented Dec 19, 2025

Uh oh!

gbaraldi commented Jan 19, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

gbaraldi commented Dec 9, 2025 •

edited

Loading

gbaraldi Dec 11, 2025 •

edited

Loading

topolarity left a comment •

edited

Loading