fix(compose): free WASM images on throw + cap pixel/raw sizes (review M2/M3)#116
Merged
Merged
Conversation
… M2/M3) Two confirmed-Medium resource-safety findings in the compose path. M2 — packages/limner-core/src/compose/photon-ops.ts: every wrapper freed its PhotonImage(s) only on the success path. A caller-triggerable throw (out-of-bounds crop, undecodable bytes) orphaned the decoded WASM buffer (multi-MB) — the server catch keeps the worker alive, so the leak is silent and accumulates across requests in a warm isolate toward the 128 MB cap. Wrapped each op in try/finally so img/out (and watermark's two images) are freed on any path. Protects both the MCP and CMA consumers (shared core). M3 — compose handler (mcp tools/compose.ts + the cma-tools twin): width/height were validated only as positive ints with no pixel-count ceiling, so a request like 100000x100000 forced a multi-GB RGBA allocation inside the isolate (OOM). And encode built a Uint8ClampedArray from caller raw.data without checking its length matched width*height*4, so the codec could read past the buffer (corrupt output) or pair a tiny buffer with huge dimensions. Added a 10 MP cap (MAX_COMPOSE_PIXELS, the documented in-isolate ceiling; oversize -> cf-images) on resize/crop/renderText/encode, and an exact raw-length check on encode. Mirrored into @limner/cma-tools for parity (D-RA-12). Tests: mocked-photon free-on-throw for all 7 ops (incl. watermark frees both); oversize-resize and encode-length-mismatch rejected cleanly in both the MCP and CMA dispatchers. Suite 677 -> 692. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Signed-off-by: Jim Vinson <jim@vinson.org>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Batch B of the max-strength review fixes — the two confirmed-Medium resource-safety findings in the compose path (both 3/3 on the adversarial pass).
M2 —
photon-opsfree WASM images on throwphoton-ops.ts freed each
PhotonImageonly on the success path. A caller-triggerable throw (out-of-bounds crop, undecodable bytes) orphaned the decoded WASM buffer; the server catch keeps the worker alive, so the leak is silent and accumulates across requests in a warm isolate toward the 128 MB cap. Every op now wraps allocation intry/finally. Shared core → protects both the MCP and CMA consumers.M3 — compose pixel cap + raw-length validation
width/heighthad no pixel-count ceiling, so100000×100000forced a multi-GB RGBA allocation → isolate OOM. Andencodebuilt aUint8ClampedArrayfrom callerraw.datawithout checkinglength === width*height*4. Added a 10 MP cap (MAX_COMPOSE_PIXELS, the documented in-isolate ceiling; oversize routes to cf-images) on resize/crop/renderText/encode, plus an exact raw-length check on encode. Mirrored into@limner/cma-toolsfor parity (D-RA-12) — the twin had the identical gap.Tests
Suite 677 → 692. Typecheck + lint clean.
🤖 Generated with Claude Code