fix: avoid deadlock on Read after WriteTo drain

[Solaris-star]

Summary

Calling io.Copy (which uses WriteTo) then Read on the same pgzip.Reader deadlocked:

rdr, _ := pgzip.NewReader(bytes.NewReader(gzipData))
io.Copy(io.Discard, rdr) // drains via WriteTo
io.ReadAll(rdr)          // hung forever

Root cause: after the final block, Read still did z.blockPool <- z.current with a nil/empty current while the pool was already full.

Fix

• If lastBlock and no remaining current, Read returns 0, io.EOF immediately
• Only return buffers with capacity to blockPool
• WriteTo never returns io.EOF (so io.Copy does not surface it)

Test plan

• go test .
• TestReadAfterWriteToNoDeadlock
• TestWriteToDoesNotReturnEOF

Fixes #39
Fixes #38

Summary by CodeRabbit

• Bug Fixes
  • Improved gzip reader end-of-stream handling to prevent potential blocking when no more data is buffered.
  • Fixed reset behavior to fully clear multistream-related internal state before re-reading headers.
  • Updated WriteTo end-of-stream behavior so callers don’t receive io.EOF.
• Tests
  • Added coverage ensuring reads after WriteTo complete do not deadlock and return io.EOF with 0 bytes.
  • Added coverage verifying WriteTo returns the correct decompressed byte count and no io.EOF error.

[klauspost]

@coderabbitai review

[coderabbitai]

✅ Action performed

Review finished.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: b6f2d2b9-607a-4744-bbab-915e230a7575

📥 Commits

Reviewing files that changed from the base of the PR and between 57d6b29 and 22ce4dd.

📒 Files selected for processing (2)

• gunzip.go
• gunzip_test.go

💤 Files with no reviewable changes (1)

• gunzip_test.go

🚧 Files skipped from review as they are similar to previous changes (1)

• gunzip.go

---

📝 Walkthrough

Walkthrough

Reader stream lifecycle handling now resets terminal state, avoids post-completion blocking, safely manages pooled buffers, and makes WriteTo return successful completion without io.EOF.

Changes

Reader stream lifecycle

Layer / File(s)	Summary
Reset and Read state guards gunzip.go, gunzip_test.go	Reader.Reset clears terminal and buffered state; Reader.Read returns io.EOF for completed streams and only pools non-nil buffers. Post-drain reads are tested for immediate EOF without deadlock.
WriteTo completion semantics gunzip.go, gunzip_test.go	Reader.WriteTo skips empty output, conditionally returns buffers to the pool, suppresses terminal io.EOF, and is tested for correct byte counts and nil completion errors.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly describes the main fix: avoiding deadlock when Read is called after WriteTo drains the reader.
Linked Issues check	✅ Passed	The changes address the deadlock and spurious EOF issues after WriteTo completes, with tests covering both behaviors for #38 and #39.
Out of Scope Changes check	✅ Passed	The modified reader logic and tests are directly tied to the linked EOF and deadlock fixes, with no unrelated changes visible.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@gunzip.go`:
- Around line 175-177: Preserve blockPool ownership on both abandonment paths in
gunzip.go: in the reset logic around lines 175-177, return z.current to
blockPool before clearing it; in the write/error handling around lines 570-584,
return read.b to blockPool before every short-write or writer-error return. Use
the existing pool-return mechanism and retain the current control flow
otherwise.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 538afa20-f79f-43d1-b69e-d4608884cc5f

📥 Commits

Reviewing files that changed from the base of the PR and between 89611d0 and 57d6b29.

📒 Files selected for processing (2)

• gunzip.go
• gunzip_test.go

[klauspost]

PTAL at the feedback and the formatting.

[Solaris-star]

Addressed the actionable feedback in 22ce4dd:

• return read.b to blockPool before both short-write and writer-error exits
• ran gofmt (including the extra blank line in gunzip_test.go)

I did not add a second Reset return because killReadAhead already returns current and clears it, avoiding a double return. Validation: go test ./... passes. PTAL.

[klauspost]

lgtm