[agarg21]

Description

minimum_versions.yml only runs on pull_request. sccache's GitHub Actions cache backend therefore stores every compiled object in a PR-scoped cache (refs/pull/<N>/merge), which no other PR can read. When we checked, the repo had 0 cache entries on main and ~24.6k PR-scoped ones. As a result every new PR's first run compiles TE from scratch: ~2.6% sccache hit rate and ~60 min median wall time, versus ~10 min per job once that PR's own cache is warm.

Running the workflow on pushes to main fills the base-branch cache scope, which PR runs can restore.

Fixes # (none filed)

Type of change

• Documentation change (change only to the documentation, either a fix or a new content)
• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Infra/Build change
• Code refactoring

Changes

• Add push: branches: [main] to .github/workflows/minimum_versions.yml (3 lines).

Measurement

Measured on a fork (agarg21/TransformerEngine), 3 fresh PRs per arm, all green:

	First run of a new PR (median)	Job-minutes per run	sccache hit rate
Before (no main scope)	63.9 min	116	2.6%
After (main warmed by one push run)	14.7 min	24	96.4%

Fork run IDs: before 39471922855, 39471924582, 39471927842; after 39496253340, 39496257361, 39496261774; warm-up push run 39482052619.

Caveats:

• Only each PR's first run benefits; later pushes to the same PR already hit that PR's own cache. That is roughly 88 of ~986 PR runs/month.
• Cost: one extra build per merge to main on GitHub-hosted runners, mostly warm after the first.
• Revert: remove the 3 lines.

Checklist:

• I have read and followed the contributing guidelines
• The functionality is complete
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes

Prepared with AI assistance (Claude) and measured on a fork. Opened as a draft per AGENTS.md until the submitter has reviewed it. Commit is DCO signed-off.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Rt4UYn6zqR9uS7DWNGEBYx

[greptile-apps]

Confidence Score: 5/5

[Low impact] The PR appears safe to merge.

Summary

Adds pushes to main as a trigger for .github/workflows/minimum_versions.yml, so builds can warm the main-branch sccache cache.

• Pushes to main now run the minimum-version checks.

agarg21 explicitly acknowledged the extra build cost as the cost of warming the main cache.

_{Reviews (1) · Last reviewed commit: "Run minimum-versions build on pushes to ..." · Reviewed by Greptile}