redefinition warnings and error on exit when using with CairoMakie

[ExpandingMan]

Whenever I import both UnicodePlots.jl and CairoMakie.jl, I get the following warnings

WARNING: redefinition of constant FT_LIB. This may fail, cause incorrect answers, or produce other errors.
WARNING: redefinition of constant VALID_FONTPATHS. This may fail, cause incorrect answers, or produce other errors.
WARNING: redefinition of constant FT_FONTS. This may fail, cause incorrect answers, or produce other errors.

(when the second package is imported, order doesn't matter).

After getting these warnings, when you exist Julia, you get

Library == CNULL. done() called before init(), or done called two times?
Stacktrace:
 [1] error(s::String)
   @ Base ./error.jl:35
 [2] ft_done()
   @ UnicodePlots.FreeTypeExt ~/.julia/packages/UnicodePlots/VGDaL/ext/FreeTypeExt.jl:421
 [3] _atexit()
   @ Base ./initdefs.jl:372

I assume there is some dependency of both packages that is behaving badly when imported twice. Anyone know where this might be coming from?

[t-bltg]

Hum, I've shamelessly copied a portion of FreeTypeAbstraction.jl in https://github.com/JuliaPlots/UnicodePlots.jl/pull/294 to reduce dependencies (user request at https://github.com/JuliaPlots/UnicodePlots.jl/issues/291).

However, the WARNING: redefinition is weird, since the code lives in https://github.com/JuliaPlots/UnicodePlots.jl/blob/master/ext/FreeTypeExt.jl now, which is a dedicated module.

Is it on 1.9.0-beta4 with package extensions - weak deps or on an earlier julia version (thus using Requires) ?

[ExpandingMan]

This is on 1.8.5.

[t-bltg]

Could you clarify what the reproducer is (with versions) ?

This doesn't give me anything, neither on 1.8.5, nor on 1.9.0-beta4 (linux):

$ julia
julia> using FreeType, FileIO  # optional, make sure `FreeTypeExt` from `UnicodePlots` will be loaded
julia> using CairoMakie
julia> using UnicodePlots  # the import order doesn't seem to matter
julia> exit(0)
$ # no warnings ?

On CairoMakie@0.10.2, UnicodePlots@3.4.0.

[ExpandingMan]

Here is a MWE, in a fresh environment, I get

◖◗ using UnicodePlots, CairoMakie
[ Info: Precompiling CairoMakie [13f3f980-e62b-5c42-98c6-ff1f3baf88f0]
WARNING: Makie.MakieLayout is deprecatedThe module `MakieLayout` has been removed and integrated into Makie, so simply replace all usage of `MakieLayout` with `Makie`.
  likely near none:1
WARNING: importing deprecated binding Makie.MakieLayout into CairoMakie.
WARNING: redefinition of constant FT_LIB. This may fail, cause incorrect answers, or produce other errors.
WARNING: redefinition of constant VALID_FONTPATHS. This may fail, cause incorrect answers, or produce other errors.
WARNING: redefinition of constant FT_FONTS. This may fail, cause incorrect answers, or produce other errors.

Manifest.log Project.log

Here are the project and manifest of the environment... I had to change the extension to .log to trick github into letting me upload them, how silly.

This is on linux btw.

[t-bltg]

Hum, still nothing on my side on 1.8.5. Anything special in your startup.jl ? Maybe Revise ?

[ExpandingMan]

Revise and OhMyREPL. My startup.jl can be found here.

Perhaps Revise is improperly trying to redefine stuff for some reason?

[t-bltg]

Right, that's Revise interfering for whatever reason:

julia> VERSION
v"1.8.5"
julia> using Revise
julia> using UnicodePlots, CairoMakie
WARNING: redefinition of constant FT_LIB. This may fail, cause incorrect answers, or produce other errors.
WARNING: redefinition of constant VALID_FONTPATHS. This may fail, cause incorrect answers, or produce other errors.
WARNING: redefinition of constant FT_FONTS. This may fail, cause incorrect answers, or produce other errors.
julia> UnicodePlots.FreeTypeExt
UnicodePlots.FreeTypeExt

But it doesn't occur on 1.9.0-beta4 (package extensions - weak deps):

julia> VERSION
v"1.9.0-beta4"
julia> using Revise
julia> using UnicodePlots, CairoMakie
julia> Base.get_extension(UnicodePlots, :FreeTypeExt)
FreeTypeExt

[ExpandingMan]

Hm... worth opening an issue with Revise maybe? I assume there must be something unusual happening, do you have an idea what might specifically be causing Revise to do this?

[t-bltg]

do you have an idea what might specifically be causing Revise to do this?

Nope no idea, I think UnicodePlots using Requires with multiple requirements is valid, and the offending code defining FT_LIB, VALID_FONTPATHS and friends in enclosed in a module, so it shouldn't happen.

Looks like __init__ is called twice somehow.

worth opening an issue with Revise maybe?

Ping @timholy if he has any clue on this (hoping he's not too busy :slightly_smiling_face: ).

[asinghvi17]

From the error statement, it seems like CairoMakie is out of date.

Could you post the version of CairoMakie and Makie for which this happened? Perhaps upgrading may help here.

[t-bltg]

CairoMakie 0.10.2 is already the latest release (Makie 0.19.2), see the .log posted by @ExpandingMan.

[ericphanson]

I'm getting crashes in a complicated setup with PythonCall in a package for which UnicodePlots is an indirect dependency which seems to point at this:

#25 63.34 WARNING: redefinition of constant FT_LIB. This may fail, cause incorrect answers, or produce other errors.
#25 63.34 WARNING: redefinition of constant VALID_FONTPATHS. This may fail, cause incorrect answers, or produce other errors.
#25 63.41 WARNING: redefinition of constant FT_FONTS. This may fail, cause incorrect answers, or produce other errors.
#25 84.61 WARNING: redefinition of constant FT_LIB. This may fail, cause incorrect answers, or produce other errors.
#25 84.61 WARNING: redefinition of constant VALID_FONTPATHS. This may fail, cause incorrect answers, or produce other errors.
#25 84.66 WARNING: redefinition of constant FT_FONTS. This may fail, cause incorrect answers, or produce other errors.
#25 95.74 Library == CNULL. done() called before init(), or done called two times?
#25 97.03 Stacktrace:
#25 97.03  [1] error(s::String)
#25 97.07    @ Base ./error.jl:35
#25 97.07  [2] ft_done()
#25 97.07    @ UnicodePlots.FreeTypeExt ~/.julia/packages/UnicodePlots/VGDaL/ext/FreeTypeExt.jl:421
#25 97.07  [3] _atexit()
#25 97.07    @ Base ./initdefs.jl:372
#25 101.9 WARNING: redefinition of constant FT_LIB. This may fail, cause incorrect answers, or produce other errors.
#25 101.9 WARNING: redefinition of constant VALID_FONTPATHS. This may fail, cause incorrect answers, or produce other errors.
#25 102.0 WARNING: redefinition of constant FT_FONTS. This may fail, cause incorrect answers, or produce other errors.
#25 113.8 Library == CNULL. done() called before init(), or done called two times?
#25 115.5 Stacktrace:
#25 115.5  [1] error(s::String)
#25 115.5    @ Base ./error.jl:35
#25 115.5  [2] ft_done()
#25 115.5    @ UnicodePlots.FreeTypeExt ~/.julia/packages/UnicodePlots/VGDaL/ext/FreeTypeExt.jl:421
#25 115.5  [3] _atexit()
#25 115.5    @ Base ./initdefs.jl:372
#25 120.2 WARNING: redefinition of constant FT_LIB. This may fail, cause incorrect answers, or produce other errors.
#25 120.2 WARNING: redefinition of constant VALID_FONTPATHS. This may fail, cause incorrect answers, or produce other errors.
#25 120.3 WARNING: redefinition of constant FT_FONTS. This may fail, cause incorrect answers, or produce other errors.
#25 131.4 Library == CNULL. done() called before init(), or done called two times?
#25 133.1 Stacktrace:
#25 133.1  [1] error(s::String)
#25 133.2    @ Base ./error.jl:35
#25 133.2  [2] ft_done()
#25 133.2    @ UnicodePlots.FreeTypeExt ~/.julia/packages/UnicodePlots/VGDaL/ext/FreeTypeExt.jl:421
#25 133.2  [3] _atexit()
#25 133.2    @ Base ./initdefs.jl:372
#25 134.2 fatal: error thrown and no exception handler available.
#25 134.2 InitError(mod=:C, error=ErrorException("no environment in the LOAD_PATH (["/root/.julia/environments/v1.8/Project.toml", "/usr/local/julia/share/julia/stdlib/v1.8"]) depends on CondaPkg"))
#25 134.3 error at ./error.jl:35
#25 134.3 _resolve_top_env at /root/.julia/packages/CondaPkg/j2Cct/src/resolve.jl:16
#25 134.3 #resolve#38 at /root/.julia/packages/CondaPkg/j2Cct/src/resolve.jl:404
#25 134.4 resolve at /root/.julia/packages/CondaPkg/j2Cct/src/resolve.jl:384
#25 134.4 envdir at /root/.julia/packages/CondaPkg/j2Cct/src/env.jl:70 [inlined]
#25 134.4 init_context at /root/.julia/packages/PythonCall/XgP8G/src/cpython/context.jl:56
#25 134.4 __init__ at /root/.julia/packages/PythonCall/XgP8G/src/cpython/CPython.jl:21
#25 134.4 jfptr___init___165614 at /deps.so (unknown line)
#25 134.4 _jl_invoke at /cache/build/default-amdci4-6/julialang/julia-release-1-dot-8/src/gf.c:2367 [inlined]
#25 134.4 ijl_apply_generic at /cache/build/default-amdci4-6/julialang/julia-release-1-dot-8/src/gf.c:2549
#25 134.4 jl_apply at /cache/build/default-amdci4-6/julialang/julia-release-1-dot-8/src/julia.h:1839 [inlined]
#25 134.4 jl_module_run_initializer at /cache/build/default-amdci4-6/julialang/julia-release-1-dot-8/src/toplevel.c:75
#25 134.4 _finish_julia_init at /cache/build/default-amdci4-6/julialang/julia-release-1-dot-8/src/init.c:765
#25 134.4 julia_init at /cache/build/default-amdci4-6/julialang/julia-release-1-dot-8/src/init.c:698
#25 134.4 jl_repl_entrypoint at /cache/build/default-amdci4-6/julialang/julia-release-1-dot-8/src/jlapi.c:713
#25 134.4 main at /cache/build/default-amdci4-6/julialang/julia-release-1-dot-8/cli/loader_exe.c:59
#25 134.4 __libc_start_main at /lib/x86_64-linux-gnu/libc.so.6 (unknown line)
#25 134.4 unknown function (ip: 0x401098)

[ericphanson]

Actually I think the real error is unrelated (InitError(mod=:C, error=ErrorException("no environment in the LOAD_PATH ...) and this issue is just scary warning spam

[t-bltg]

I've opened an upstream report at https://github.com/timholy/Revise.jl/issues/736 with a minimal reproducer, since I believe this is an upstream Revise bug.