Precompile Markdown parsing AND printing

[PallHaraldsson]

Improvement on top of #55715, replacing it, to also precompile printing. And then also fixing parsing and using regression for #55706.

[PallHaraldsson]

[This shouldn't need a full CI, since a trivial tested add-on.]

This is likely the better backport target for 1.11, assuming it only adds to pkgimage (i.e. not the sysimage), not concerned about bloating it, a bit.

This could precompile a bit more for stdout, not just stderr, but unsure how to do that, also likely not a large improvement. It also misses some tags, for images, not as important, maybe easy to add, but I don't want this to hold up releasing 1.11. Can always be added later...

[PallHaraldsson]

Julia support the CommonMark extension: https://www.markdownguide.org/extended-syntax/ but is it the only one, e.g. no R Markdown allowed? The parser will not fail at any extension, just ignore it, so might be ok to include here, for later compatibility?

I was looking into what basic tags to add here, and what more, maybe, I see we likely do not support latest 2024 CommonMark https://spec.commonmark.org/0.31.2/changes.html since I found in the code: https://spec.commonmark.org/0.19/#absolute-uri

[KristofferC]

Note that this will precompile many things for

julia> typeof(devnull)
Base.DevNull

which is not the type that will typically be used when printing things.

[PallHaraldsson]

This is now done.

You can't simply skip devnull, because then it would show, and I'm not sure how to avoid that differently, than I do now.

[PallHaraldsson]

I think this is now done for the 1.11 backports. Only question is the precompile I got for Markdown (probably harmless to have here; or skip it): precompile(Tuple{typeof(Base.:(^)), String, Int64})

[PallHaraldsson]

Bump. This was ok for me locally, the CI failure is most likely unrelated, and this can and should be merged, and ideally go into 1.11, since a small better change and this line seems not to be about me:

julia/test/runtests.jl

Line 100 in 7a76e32

cd(@__DIR__) do

    Pkg          |           1      1  7m05.9s
    FAILURE
The global RNG seed was 0x6aeb1dca8020ba26586d7a4ba73abb95.
Error in testset Pkg:
Error During Test at none:1
  Got exception outside of a @test
  ProcessExitedException(2)
  Stacktrace:
    [1] try_yieldto(undo::typeof(Base.ensure_rescheduled))
      @ Base ./task.jl:1111
    [2] wait()
      @ Base ./task.jl:1175
    [3] wait(c::Base.GenericCondition{ReentrantLock}; first::Bool)
      @ Base ./condition.jl:141
    [4] wait
      @ ./condition.jl:136 [inlined]
    [5] take_buffered(c::Channel{Any})
      @ Base ./channels.jl:493
    [6] take!(c::Channel{Any})
      @ Base ./channels.jl:487
    [7] take!(::Distributed.RemoteValue)
      @ Distributed /cache/build/tester-amdci5-15/julialang/julia-master/julia-628169b274/share/julia/stdlib/v1.12/Distributed/src/remotecall.jl:726
    [8] remotecall_fetch(::Function, ::Distributed.Worker, ::String, ::Vararg{String}; kwargs::@Kwargs{seed::UInt128})
      @ Distributed /cache/build/tester-amdci5-15/julialang/julia-master/julia-628169b274/share/julia/stdlib/v1.12/Distributed/src/remotecall.jl:461
    [9] remotecall_fetch(::Function, ::Int64, ::String, ::Vararg{String}; kwargs::@Kwargs{seed::UInt128})
      @ Distributed /cache/build/tester-amdci5-15/julialang/julia-master/julia-628169b274/share/julia/stdlib/v1.12/Distributed/src/remotecall.jl:492
   [10] (::var"#46#47"{Vector{Task}, var"#print_testworker_errored#39"{ReentrantLock, Int64, Int64}, var"#print_testworker_stats#36"{ReentrantLock, Int64, Int64, Int64, Int64, Int64, Int64}, Vector{Any}, Dict{String, DateTime}})()
      @ Main /cache/build/tester-amdci5-15/julialang/julia-master/julia-628169b274/share/julia/test/runtests.jl:264
ERROR: LoadError: Test run finished with errors
in expression starting at /cache/build/tester-amdci5-15/julialang/julia-master/julia-628169b274/share/julia/test/runtests.jl:100
ERROR: A test has failed. Please submit a bug report (https://github.com/JuliaLang/julia/issues)

[PallHaraldsson]

Bump.

False alarm CI is a problem for PRs getting accepted I think (a bit too common IMHO, then I miss out on some real CI failures):
Pkg (2) | failed at 2024-09-15T00:55:04.032

[PallHaraldsson]

This can be merged as is. I'm trying a rebase though (for the first time), I think that's the solution to CI false alarms. I meant to do rebase, after button seemed not responding, think I chose the option above, "to branch"(?) by accident, not even sure what it does or if I made worse...

[fredrikekre]

Any timings before/after?

[PallHaraldsson]

E.g.:

julia> @time println(md"# H1")
# H1

  0.000095 seconds (13 allocations: 296 bytes)

before:
julia> @time using Markdown
  0.049139 seconds (16.95 k allocations: 1.171 MiB)

julia> @time println(md"# H1")
# H1

  0.103901 seconds (32.81 k allocations: 1.701 MiB, 99.73% compilation time)

[fredrikekre]

Ok, and those 0.1 seconds are annoying/noticable when?

[PallHaraldsson]

This doesn't affect the sysimage size, so I think precompilng as much as possible ok here?

I had already fixed the major issue I had with using Markdown being slow, I just thought being more complete would be helpful, in a follow-up PR (I dislike seeing those "precompilation time", they add up..., it's annoying when looking for a root cause which is the real cause). There might be others slower, just showing some before and after, since asked for.