Windows issue with with partition dependency compiler when `MIX_OS_DEPS_COMPILE_PARTITION_COUNT > 1`

[infinityfye]

Elixir and Erlang/OTP versions

Erlang/OTP 28 [erts-16.3] [source] [64-bit] [smp:16:16] [ds:16:16:10] [async-threads:1] [jit:ns]

Elixir 1.19.5 (compiled with Erlang/OTP 28)

Operating system

Windows 11 Pro (25H2) x86_64

Current behavior

In a fairly new phoenix project, at some point after adding dependencies an issue shows up when during the dependency compilation. That is whether manually triggered via mix deps.compile or when running any other command that triggers compilation:

> mix deps.compile
mix deps.compile running across 8 OS processes
** (Mix) could not start partition dependency compiler, no connection made to TCP port: {:error, :timeout}

1 >The command line is too long.

2 >The command line is too long.

3 >The command line is too long.

4 >The command line is too long.

5 >The command line is too long.

6 >The command line is too long.

7 >The command line is too long.

8 >The command line is too long.

This is resolved by setting MIX_OS_DEPS_COMPILE_PARTITION_COUNT to 1. This issue doesn't happen on WSL either.

I am providing my deps function here if someone can try and replicate this:

defp deps do
  [
    # documentation
    {:ex_doc, "~> 0.39", only: :dev, runtime: false, warn_if_outdated: true},

    # database
    {:postgrex, ">= 0.0.0"},
    {:ecto_sql, "~> 3.13"},

    # clustering
    {:libcluster, "~> 3.5"},
    {:horde, "~> 0.10"},

    # http server
    {:bandit, "~> 1.5"},

    # http client
    {:req, "~> 0.5"},

    # telemetry
    {:telemetry_metrics, "~> 1.0"},
    {:telemetry_poller, "~> 1.0"},

    # errors
    {:tower, "~> 0.8"},
    {:tower_error_tracker, "~> 0.3.8"},

    # phoenix
    {:phoenix, "~> 1.8"},
    {:phoenix_ecto, "~> 4.5"},
    {:phoenix_html, "~> 4.1"},
    {:phoenix_live_reload, "~> 1.2", only: :dev},
    {:phoenix_live_view, "~> 1.1.0"},
    {:esbuild, "~> 0.10", runtime: Mix.env() == :dev},
    {:tailwind, "~> 0.4", runtime: Mix.env() == :dev},
    {:heroicons,
      github: "tailwindlabs/heroicons", tag: "v2.2.0", sparse: "optimized", app: false, compile: false, depth: 1},

    # email
    {:swoosh, "~> 1.16"},

    # config
    {:dotenvy, "~> 1.1"},

    # oban
    {:oban, "~> 2.20"},
    {:oban_web, "~> 2.11"},

    # utils
    {:uniq, "~> 0.6"},
    {:remote_ip, "~> 1.2"},
    {:phoenix_live_dashboard, "~> 0.8"},
    {:ecto_psql_extras, "~> 0.8"},
    {:live_debugger, "~> 0.5", only: :dev},
    {:igniter, "~> 0.7", only: [:dev, :test]},

    # rich text and patches
    {:delta, "~> 0.4"},
    {:jsonpatch, "~> 2.0"},

    # i18n
    {:gettext, "~> 1.0"},

    # testing, quality, static analysis
    {:lazy_html, ">= 0.1.0", only: :test},
    {:phoenix_test, "~> 0.9", only: :test, runtime: false},
    {:credo, "~> 1.7", only: [:dev, :test], runtime: false},
    {:dialyxir, "~> 1.4", only: [:dev, :test], runtime: false},
    {:sobelow, "~> 0.14.1", only: [:dev, :test], runtime: false, warn_if_outdated: true},
    {:benchee, "~> 1.0", only: [:dev, :test]},
    {:benchee_markdown, "~> 0.3", only: [:dev, :test]},
    {:benchee_html, "~> 1.0", only: [:dev, :test]},

    # ai stuff
    {:usage_rules, "~> 1.1", only: [:dev]},
    {:tidewave, "~> 0.5", only: [:dev]}
  ]
end

Expected behavior

When MIX_OS_DEPS_COMPILE_PARTITION_COUNT > 1 it should work on windows as it does on other platforms.

I hope I've given enough information to be of any use. Please let me know if there is anything I can do to make this issue more reproducible.

[josevalim]

We pass some metadata via CLI args. If we are running into limits on Windows, we probably need to pass it through a file or using another mechanism.