Repository navigation
fix: guard against patching binary files - #9
Merged
Merged
Conversation
semantic-release derives the published version from git tags at CI time, so the committed version field was diverging (0.1.3 vs published 0.3.0). Set it to 0.0.0 to signal it's a placeholder managed by release tooling. Co-Authored-By: Claude Code <noreply@anthropic.com>
prepareOperation read files as UTF-8, decoding binary into replacement characters so a patch could match garbage or silently corrupt the file on write. Read as Buffer, reject on a null byte with an EBINARY code before any write, then decode. Delete hunks never read content and are unaffected. Co-Authored-By: Claude Code <noreply@anthropic.com>
This reverts commit 8c0cbc3.
Co-Authored-By: Claude Code <noreply@anthropic.com>
Contributor
Reviewer's GuideAdds an early null-byte guard for apply_patch updates and moves so binary files cannot be decoded or corrupted, while preserving deletion semantics and reporting structured EBINARY failures without misleading previews. Sequence diagram for binary-safe apply_patch updatessequenceDiagram
participant ApplyPatch
participant Preview as createPatchPreview
participant Prepare as prepareOperation
participant Guard as readTextFileRejectingBinary
participant FS as FileSystem
ApplyPatch->>Preview: createPatchPreview(cwd, hunks)
Preview->>Guard: readTextFileRejectingBinary(absolutePath, filePath)
Guard->>FS: readFile(absolutePath)
alt null byte found
Guard-->>Preview: Error with code EBINARY
Preview-->>ApplyPatch: Structured failure with file path
else text file
Guard-->>Preview: UTF-8 content
Preview->>Prepare: prepareOperation(cwd, hunk)
Prepare->>Guard: readTextFileRejectingBinary(absolutePath, filePath)
Guard-->>Prepare: UTF-8 content
Prepare-->>ApplyPatch: Safe patch operation
end
Flow diagram for binary-safe apply_patch operationsflowchart TD
A[apply_patch update or move] --> B{Operation type}
B -->|delete| C[Preserve existing deletion behavior]
B -->|update or move| D[readTextFileRejectingBinary]
D --> E[readFile]
E --> F{Contains null byte?}
F -->|yes| G[Reject with EBINARY]
G --> H[No decode, write, move, or text preview]
F -->|no| I[Decode as UTF-8]
I --> J[Create preview and prepare write or move]
File-Level Changes
Tips and commandsInteracting with Sourcery
Customizing Your ExperienceAccess your dashboard to:
Getting Help
|
paulpham157
marked this pull request as ready for review
September 15, 2026 09:15
![sourcery-ai[bot]](https://avatars.githubusercontent.com/in/48477?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Hey - I've found 1 issue
Prompt for AI Agents
Please address the comments from this code review:
## Individual Comments
### Comment 1
<location path="src/index.ts" line_range="1201-1205" />
<code_context>
+ // bytes into replacement characters, so a patch could either match
+ // garbage or silently corrupt the file on write. Reject early with a
+ // distinct code instead of letting apply_patch guess at binary content.
+ const raw = await readFile(absolutePath);
+ if (raw.includes(0)) {
+ throw Object.assign(new Error(`Refusing to patch binary file: ${hunk.filePath}`), { code: "EBINARY" });
+ }
+ const currentContent = raw.toString("utf-8");
const result =
hunk.chunks.length === 0
</code_context>
<issue_to_address>
**issue (broader_impact):** The tool's initial preview path still reads existing update targets with the UTF-8 string overload before `prepareOperation` runs, so a binary file is decoded into replacement characters and can be shown as a misleading text diff before the later `EBINARY` rejection. This violates the change's stated guarantee that binary content is rejected before decoding.
**Triggers:** When the patch is submitted through the `apply_patch` tool and the target contains invalid UTF-8 or binary bytes.
**Suggested fix:** Make preview generation use the same raw-byte null-byte check, or skip text preview for binary targets and let the operation preparation report `EBINARY`.
</issue_to_address>Sourcery assessment
Approval pending. 1 finding to address first.
Blocking findings: src/index.ts:1205
Reuse the null-byte guard when generating update previews so apply_patch falls back to a progress-only pending update rather than decoding binary content. Keep deletes unguarded and cover binary update, move, delete, and preview behavior. Co-Authored-By: Claude Code <noreply@anthropic.com>
paulpham157
marked this pull request as draft
September 15, 2026 09:48
paulpham157
marked this pull request as ready for review
September 15, 2026 09:49
Contributor
There was a problem hiding this comment.
Hey - I've found 1 issue
Prompt for AI Agents
Please address the comments from this code review:
## Individual Comments
### Comment 1
<location path="src/index.ts" line_range="919-922" />
<code_context>
continue;
}
- const oldContent = await readFile(absolutePath, "utf-8");
+ const oldContent = await readTextFileRejectingBinary(absolutePath, hunk.filePath);
const newContent =
hunk.chunks.length === 0 ? oldContent : replaceChunks(oldContent, hunk.filePath, hunk.chunks).content;
</code_context>
<issue_to_address>
**issue (broader_impact):** Binary delete operations still read the file with `readFile(..., "utf-8")` in the delete branch, so the pending and final tool results include replacement-character text previews for binary contents even though binary update previews are suppressed. This produces the misleading preview the change is intended to prevent.
**Triggers:** When `apply_patch` deletes a binary file through the tool interface.
**Suggested fix:** Read binary deletes as bytes and omit the preview, or detect the null byte in the delete preview path and fall back to the progress-only result.
</issue_to_address>Sourcery assessment
Approval pending. 1 finding to address first.
Blocking findings: src/index.ts:922
Comment on lines
-919
to
922
| const oldContent = await readFile(absolutePath, "utf-8"); | ||
| const oldContent = await readTextFileRejectingBinary(absolutePath, hunk.filePath); | ||
| const newContent = | ||
| hunk.chunks.length === 0 ? oldContent : replaceChunks(oldContent, hunk.filePath, hunk.chunks).content; | ||
| if (hunk.movePath) { |
Contributor
There was a problem hiding this comment.
issue (broader_impact): Binary delete operations still read the file with readFile(..., "utf-8") in the delete branch, so the pending and final tool results include replacement-character text previews for binary contents even though binary update previews are suppressed. This produces the misleading preview the change is intended to prevent.
Triggers: When apply_patch deletes a binary file through the tool interface.
Suggested fix: Read binary deletes as bytes and omit the preview, or detect the null byte in the delete preview path and fall back to the progress-only result.
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.
Summary
Reject
apply_patchupdates to files containing a null byte before decoding or writing them. The failure is reported withEBINARYand includes the patched file path, preventing UTF-8 replacement-character corruption.Regression coverage verifies structured
EBINARYfailures, byte-for-byte preservation, binary move rejection, binary deletion, and that pending/final tool results omit misleading text previews.Before opening this PR
Verification
npm run check(typecheck + biome)npm test(127 tests)npm pack --dry-run(release sanity)git diff --check main...HEADTested revision:
7143f90Node / Pi versions: Node
v26.8.1; Pi0.85.1Local OS: macOS (Darwin 25.6.0)
Live Pi results
Report or evidence: Local deterministic regression test covers null-byte rejection and exact byte preservation.
Transport evidence if applicable (grammar / JSON / unverified): Unverified.
Failures, missing coverage, or N/A reason: This is a filesystem/path safety change. Per CONTRIBUTION.md, required developer-run live CRUD and CI on Ubuntu, macOS, and Windows remain outstanding. This PR stays draft until the live tests and CI complete, or a maintainer records a waiver.
apply_patch impact
EBINARY; no breaking migration.🤖 Generated with Claude Code
Summary by Sourcery
Guard apply_patch against binary file updates while preserving safe deletion behavior and accurate tool results.
Bug Fixes:
Enhancements:
Tests: