Skip to content

Stills of a 2D composition render into a folder that doesn't exist yet, as before DR2 - #91

Merged
shinyobjectz merged 2 commits into
shinyobjectz:mainfrom
JimGHTB12:stills-manifest-folder
Nov 10, 2026
Merged

shinyobjectz merged 2 commits into
shinyobjectz:mainfrom
JimGHTB12:stills-manifest-folder

Conversation

@JimGHTB12

@JimGHTB12 JimGHTB12 commented Nov 10, 2026 •

Copy link
Copy Markdown
Contributor

Since bb0a176 (DR2), stills of a 2D composition fail when the output folder doesn't exist yet:

$ raea render edit.tsx --at 0 -o new-folder
REFUSED  new-folder\edit.render.json: The system cannot find the path specified. (os error 3)

Nothing is drawn. It works into a folder that is already there, and for a composition with 3D.

Why

DR2 writes every render's manifest beside its output, DIR/NAME.render.json for a folder of frames. The 2D stills path calls write_manifest before raea_runtime_encode::render makes the folder and writes the frames, so the write has nowhere to go. Before DR2 there was no manifest unless --manifest asked for one. The 3D path makes the folder first, so it was not hit.

Change

write_manifest makes the manifest's folder before it writes. raea-runtime-export is at its 2,000-line limit, so the change rewrites three lines and adds none.

Two commits. The first chained the write after the folder with a closure that tools/dupes.sh reads as a copy of one in apps/check (57 near-duplicate groups against 56), and the gates job failed on it. The second makes the folder in a statement of its own.

Testing

Two of the repo's own benches render stills into a fresh temporary folder, so they are the test:

main this branch
bench/at-0/check.sh FAIL: the --at 0 render failed (the message above) PASS
bench/svg-css-transform/check.sh FAIL sheet: the render failed (the same) PASS

Run on Windows 11 with a release build of each (cargo build --release -p raea-cli --no-default-features), main at bb0a176. tools/size.sh, layers.sh, names.sh, one-version.sh and no-void.sh pass on the branch, and cargo dupes counts the same groups as on main. The same build also renders into a nested new folder, and with --manifest pointing into a folder that doesn't exist.

Not run: macOS and Linux, though std::fs::write into a missing folder fails the same way on both. cargo shear isn't installed here.

The first commit was reviewed by GPT-6.1 Sol (high effort), a different model family from the one that wrote it: one round, no findings. The second commit has not been reviewed.

Created with Claude Fable 5.2 in T3 Code (Claude Code).

🤖 Generated with Claude Code

…t, as before DR2

Since bb0a176 every render writes its manifest beside its output, DIR/NAME.render.json for a folder of frames. A 2D
composition's stills write the manifest before raea-render makes the folder and its frames, so with a DIR that
didn't exist the render stopped before drawing anything:

    raea render edit.tsx --at 0 -o new-folder
    REFUSED  new-folder/edit.render.json: The system cannot find the path specified. (os error 3)

It worked into a folder that was already there, and for a composition with 3D, whose path makes the folder first.
write_manifest now makes the manifest's folder before it writes.

bench/at-0/check.sh and bench/svg-css-transform/check.sh render stills into a fresh temporary folder: both fail on
main with the message above, and both pass with this (Windows 11).

Co-Authored-By: Claude <noreply@anthropic.com>
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Nov 10, 2026 •

Copy link
Copy Markdown
Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-11-10T22:00:08.564681Z 22cac9c PR opened
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

The first form chained the write after the folder with `.and_then(|()| std::fs::write(&path, m))`, which
tools/dupes.sh reads as a copy of apps/check's hook.rs:105 (57 near-duplicate groups against the 56 allowed).
The folder is now made first, in its own statement, and the write is as it was. The crate is still at its
2,000 lines: the manifest's JSON is written where it is built.

Co-Authored-By: Claude <noreply@anthropic.com>
@shinyobjectz
shinyobjectz merged commit 38684c9 into shinyobjectz:main Nov 10, 2026
4 checks passed
@shinyobjectz

Copy link
Copy Markdown
Owner

Merged, thanks: a clear cause, and two existing benches as the test.

shinyobjectz added a commit that referenced this pull request Nov 16, 2026
…y (80067cc, pushed and merged), with #91's characters guide and #92's reference/feed.md (plan_close, raea-feed #61), whose stale copy failed every lane's L0 and L1
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.

2 participants