More parallelism

[grahamc]

Motivation

I drafted this when I thought CI would stall again, but then didn't need it.

Context

Summary by CodeRabbit

• Chores
  • Expanded build matrix testing by increasing granularity of nix configuration coverage, changing from range-based groupings to individual configuration entries. This ensures more comprehensive testing across all supported build configurations.

[coderabbitai]

📝 Walkthrough

Walkthrough

The flake regression matrix glob patterns in the GitHub Actions build workflow were refactored from range-based groupings to explicit single-character patterns, expanding from four range patterns to separate patterns for each digit (0-9) and letter (a-z), affecting which nix configurations are included in the build matrix.

Changes

Cohort / File(s)	Summary
GitHub Actions Workflow Configuration .github/workflows/build.yml	Expanded flake regression matrix glob patterns from range-based groupings ([0-d]*, [e-l]*, [n-r]*, [s-z]*) to explicit single-character entries ([0]* through [9]*, [a]* through [z]*), broadening the set of nix configurations included in the matrix.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

Poem

🐰 Whisker-twitching with glee,
Each letter and digit now runs wild and free,
From ranges now explicit, the workflows align,
A rabbit's delight—configuration divine! ✨

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 inconclusive)

Check name	Status	Explanation	Resolution
Title check	❓ Inconclusive	The title 'More parallelism' is vague and does not clearly describe the specific change made to the workflow matrix configuration.	Consider a more specific title that describes the actual change, such as 'Expand CI matrix with per-character glob patterns' or 'Increase matrix parallelism by enumerating character patterns'.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch push-vpznklyquurl

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[github-actions]

🚀 Deployed on https://699d32f561569ac7d610d7cd--determinate-nix.netlify.app

[coderabbitai]

🧹 Nitpick comments (1)

.github/workflows/build.yml (1)

134-170: 7× increase in matrix job count (15 → 108) — confirm cost/quota implications.

Replacing 5 range globs with 36 single-character globs × 3 nix_config values yields 108 concurrent jobs per trigger, up from 15. A few things worth confirming:

1. CI cost / minute quota — runner-minutes consumed will scale proportionally. Given each job can run up to 60 minutes (inherited from the flake_regressions job has no explicit timeout, so it inherits none — unlimited), the worst-case total is unbounded.
2. Runner availability — inputs.runner is the target; confirm the pool is large enough to absorb 108 parallel jobs without queuing that defeats the parallelism goal.
3. Intentionality — the PR description notes the changes were drafted preemptively and "ultimately did not need the changes." If this is still true, landing 108-job matrices on every qualifying trigger may be unnecessary overhead.

Also worth noting: each job still runs with PARALLEL: '-P 1' (line 190), so parallelism within each job is serialized — the only gain is cross-job parallelism. The previously-commented-out adaptive logic (-P 50% for non-eval-cores configs) at line 189 could compound this further if ever re-enabled.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In @.github/workflows/build.yml around lines 134 - 170, The matrix expansion
replaced compact range globs with 36 single-character globs causing nix_config ×
those entries to balloon the flake_regressions matrix from 15 to 108 jobs;
revert to using range-style globs (e.g., the previous 5 range patterns) or
otherwise reduce the glob list to match the original cardinality, confirm with
the CI owners that inputs.runner can handle 108 simultaneous jobs and update
job-level protections (set a timeout or concurrency cap on the flake_regressions
job), and consider restoring adaptive per-job parallelism (the commented `-P
50%` logic and current `PARALLEL: '-P 1'`) so each job doesn’t serialize
unnecessarily when the matrix size increases.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Nitpick comments:
In @.github/workflows/build.yml:
- Around line 134-170: The matrix expansion replaced compact range globs with 36
single-character globs causing nix_config × those entries to balloon the
flake_regressions matrix from 15 to 108 jobs; revert to using range-style globs
(e.g., the previous 5 range patterns) or otherwise reduce the glob list to match
the original cardinality, confirm with the CI owners that inputs.runner can
handle 108 simultaneous jobs and update job-level protections (set a timeout or
concurrency cap on the flake_regressions job), and consider restoring adaptive
per-job parallelism (the commented `-P 50%` logic and current `PARALLEL: '-P
1'`) so each job doesn’t serialize unnecessarily when the matrix size increases.

---

ℹ️ Review info

Configuration used: defaults

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between 628d55c and e69b425.

📒 Files selected for processing (1)

• .github/workflows/build.yml