Skip to content

User-manual Docker page documents a non-existent ghcr.io image, variants, compose file and builder/runtime split (CR-A16-006) #451

Open

Description

@cemililik

Wave 1 · theme D1-8: Docs describing features that do not exist (doc side of code findings) · highest severity High · 1 finding(s) · effort medium

Summary

An operator follows the manual. docker pull fails, and the multi-GPU torchrun recipe exits at argument parsing.

Primary finding

CR-A16-006 — User-manual Docker page documents a non-existent ghcr.io image, variants, compose file and builder/runtime split

severity High · reachability default · intent contract-violation · confidence confirmed · independently confirmed · not reproduced (code trace) · already tracked: KNOWN-OPEN C31-DOCKER-RUNBOOK — docs/roadmap/phase-16-trust-surface-hardening.md:689-696

What happens. docs/usermanuals/{en,tr}/operations/docker.md:12-34,36-77,90 (and installation.md:103-104) instruct docker pull ghcr.io/hodetech/forgelm:latest with latest/slim/airgap variants, show a different compose file and describe a builder stage copying artefacts, while README.md:182 says no image is published, the only Dockerfile is CUDA-only with cumulative stages, and commands prefix 'forgelm' after an ENTRYPOINT that is already forgelm.

Where.

Evidence.

  • Static comparison with docker-compose.yaml and Dockerfile.
  • Static.

Recommended fix. Rewrite the page around the shipped Dockerfile and compose file, and use --entrypoint torchrun for the multi-GPU recipe.

Acceptance criteria

  • A regression test reproduces the failure scenario and asserts the corrected behaviour.
  • Documentation listed above is corrected in English and in its Turkish mirror in the same PR.
  • Every finding in this issue is resolved or explicitly closed with a reason.

Found by the October 2026 ForgeLM code and documentation review against commit cab2947. Code links point at that commit; line numbers may have moved since. Finding IDs (CR-…, DOC-…, TECH-…) are stable references for this backlog.

Metadata

Metadata

Assignees

No one assigned

    Labels

    area: packaging-ciCode area: packaging-cibacklog: knownAlready tracked in docs/roadmapblockedWaits for another issue (see Blocked by); set and removed automatically, never by handbugSomething isn't workingseverity: highWrong behaviour or a false claim users rely onsource: review-2026-10Found by the October 2026 code and documentation reviewwave: 1Roadmap wave 1: high severity, small effort

    Type

    No type

    Projects

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions