Skip to content

feat(R-CMD-check): expose system-deps and pandoc to callers - #31

Merged
eliotmcintire merged 2 commits into
mainfrom
feat/system-deps-input
Sep 1, 2026
Merged

feat(R-CMD-check): expose system-deps and pandoc to callers#31
eliotmcintire merged 2 commits into
mainfrom
feat/system-deps-input

Conversation

@eliotmcintire

Copy link
Copy Markdown
Contributor

setup-r-deps has had both switches since #24, but R-CMD-check.yaml neither exposed nor passed them — so every caller installs the geospatial stack whether or not it can use it.

The cost

Six packages declare no spatial dependency at all — no terra, sf, sp, raster:

fpCompare, Require, pedev, peutils, SpaDES.install, SpaDES.addins

Each was pulling ~3GB of Homebrew on its macOS leg — llvm 1.9GB, aws-sdk-cpp 640MB, boost, gcc — measured at ~200s, at the 10x macOS billing rate, plus an apt install on every Linux leg, for libraries they never load.

Usage

jobs:
  R-CMD-check:
    uses: PredictiveEcology/actions/.github/workflows/R-CMD-check.yaml@main
    with:
      system-deps: false   # pure-R package
      pandoc: false        # no vignettes / Rmd

Both default to true, so no existing caller changes behaviour.

Why it matters now

The org concurrency budget is shared across all repos and is currently the binding constraint. This reclaims runner time without touching what any package actually tests — the macOS legs of six repos stop doing several minutes of work each that cannot affect the result.

🤖 Generated with Claude Code

https://claude.ai/code/session_012DVjmY3im9Xak7tXLSMGCa

eliotmcintire and others added 2 commits August 31, 2026 17:24
setup-r-deps has had both switches since it was extracted, but the reusable
workflow neither exposed nor passed them, so every caller installed the
geospatial stack and pandoc whether or not it could use them.

Six packages in the org declare no spatial dependency at all -- fpCompare,
Require, pedev, peutils, SpaDES.install, SpaDES.addins -- and each was pulling
~3GB of Homebrew (llvm, gcc, boost, aws-sdk-cpp, ~200s) on its macOS leg at the
10x macOS billing rate, plus an apt install on every Linux leg, for libraries
they never load.

Both default to true, so no existing caller changes behaviour.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012DVjmY3im9Xak7tXLSMGCa
test-coverage does not call setup-r-deps -- it still carries its own inline
pandoc and apt steps -- so the inputs are wired directly onto those steps
rather than passed through. Both default to true; nothing changes for existing
callers.

Smaller payoff than on R-CMD-check: this job is ubuntu-only, so it saves an apt
install rather than the ~3GB macOS Homebrew pull.

Also worth noting the comment above its spatial block is now stale: it warns
that install-spatial-deps uses the ubuntugis PPA, which #24 removed.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012DVjmY3im9Xak7tXLSMGCa
@eliotmcintire
eliotmcintire merged commit c6e7bbb into main Sep 1, 2026
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant