Add self contained setup-rebar.sh for download and setup

[ttschampel]

Summary

Adds downloadable setup-rebar.sh script and corresponding README instruction update

Contracts

• Implements: CONTRACT:___ (new implementation)
• Modifies behavior of: CONTRACT:___ (check contract still holds)
• New contract: architecture/CONTRACT-___ (requires plan mode review)
• No contract impact (refactor, docs, tests only)

Checklist

• Every new/modified source file has a CONTRACT: header comment
• Contract references point to existing files (./scripts/check-contract-refs.sh)
• No untracked TODO: comments (./scripts/check-todos.sh)
• Tests pass at T2+ (package-level or higher)
• Docs updated if behavior changed (QUICKCONTEXT, TODO, relevant READMEs)

Test Plan

• T1 — targeted test(s): ___
• T2 — package suite: ___
• T3+ — cross-package / full suite

[damorris25]

LGTM

[willackerly]

Thanks Tim — really appreciate the signal here. The curl | bash one-liner is a real UX gap and your PR is what pushed me to close it.

I'm going to land a thinner version on v3.0.0-alpha rather than merge this directly: #2. The reason is that rebar's Go CLI already has rebar new, rebar adopt, bin/install, and ask init as the canonical onboarding path, and I want the installer to be a shim over those rather than a parallel codepath — the latter would drift the moment any of them change. The version in #2:

• clones rebar to ~/.rebar, runs bin/install, and exits pointing at rebar new <name> (no template-splatting into $PWD)
• skips auto-installing Homebrew/bash5 — bin/install already does the right "report missing deps" thing without the sudo prompt mid-curl
• ~120 lines instead of 462, no parallel bootstrap logic to keep in sync

If you'd like a worked example: backporting #2 from v3.0.0-alpha to main once alpha lands is yours for the taking — that's a clean small PR against the canonical workflow.

Either way, would love your eyes on #2 — does the shim cover your original need? Anything that felt obvious in your version that I dropped and shouldn't have?

Thanks again for filing this; the install ergonomics are clearly worth fixing and your PR is what surfaced it.

[willackerly]

Quick follow-up: #2 is merged into v3.0.0-alpha, and I've cut a v3.0.1-alpha branch with the version bump applied — that's the canonical place to look at the shim now:

• Browse: https://github.com/willackerly/rebar/tree/v3.0.1-alpha
• Script: https://github.com/willackerly/rebar/blob/v3.0.1-alpha/setup-rebar.sh
• Try it (curl|bash):

  curl -fsSL https://raw.githubusercontent.com/willackerly/rebar/v3.0.1-alpha/setup-rebar.sh | bash

If you run it end-to-end, would love to know:

1. Does it cover what your version was solving for you?
2. Anything that felt obvious in your version that I dropped and probably shouldn't have?
3. The dep-handling difference is the biggest delta — your script auto-brew install'd missing deps; mine just reports them via bin/install. Was the auto-install important to your workflow, or fine to leave as user responsibility?

[ttschampel]

Tried out https://github.com/willackerly/rebar/blob/v3.0.1-alpha/setup-rebar.sh

It worked well for me. I'm on the fence about how important auto installing dependences is vs just reporting them. The new report missing is a big improvement.

[ttschampel]

I think there are a few bit missing from the install.

• contract registry script did not install
• multi agent files did not install
• looks like some rebar audit checks checks for file existence for scripts which aren't present; the are present in the rebar installation directory though