Uh oh!
There was an error while loading. Please reload this page.
[linter-miner] feat(linters): add httprespbodyclose linter - #42816
Conversation
Add a new custom Go analysis linter that flags HTTP response Body.Close() calls that are made directly (without defer), which risks resource leaks when the function returns early due to an error. The linter tracks *http.Response variables assigned from HTTP calls (using type information) and reports when resp.Body.Close() is called without defer. Files added: - pkg/linters/httprespbodyclose/httprespbodyclose.go - pkg/linters/httprespbodyclose/httprespbodyclose_test.go - pkg/linters/httprespbodyclose/testdata/src/httprespbodyclose/httprespbodyclose.go Also registers the new analyzer in cmd/linters/main.go. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
🧠 Matt Pocock Skills Reviewer has completed the skills-based review. ✅ |
✅ Design Decision Gate 🏗️ completed the design decision gate check. |
✅ Test Quality Sentinel completed test quality analysis. |
There was a problem hiding this comment.
Pull request overview
Adds a new custom go/analysis linter (httprespbodyclose) to gh-aw’s in-repo linter suite, intended to prevent HTTP response-body leaks by flagging non-deferred resp.Body.Close() calls.
Changes:
- Introduces
pkg/linters/httprespbodycloseanalyzer and hooks it into thecmd/lintersmultichecker. - Adds analysistest-based unit tests + testdata fixtures for the new analyzer.
- Extends the linter runner (
cmd/linters/main.go) to include the new analyzer in the default linter set.
Show a summary per file
| File | Description |
|---|---|
pkg/linters/httprespbodyclose/httprespbodyclose.go | New analyzer implementation that tracks *http.Response variables and detects non-deferred Body.Close() calls. |
pkg/linters/httprespbodyclose/httprespbodyclose_test.go | Adds analysistest harness for the new analyzer. |
pkg/linters/httprespbodyclose/testdata/src/httprespbodyclose/httprespbodyclose.go | Test fixtures for expected diagnostics and non-diagnostic cases. |
cmd/linters/main.go | Registers httprespbodyclose.Analyzer in the default multichecker set. |
Review details
Tip
Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
- Files reviewed: 4/4 changed files
- Comments generated: 3
- Review effort level: Low
| fmterrorfnoverbs.Analyzer, | ||
| hardcodedfilepath.Analyzer, | ||
| httpnoctx.Analyzer, | ||
| httprespbodyclose.Analyzer, | ||
| httpstatuscode.Analyzer, |
| "github.com/github/gh-aw/pkg/linters/hardcodedfilepath" | ||
| "github.com/github/gh-aw/pkg/linters/httpnoctx" | ||
| "github.com/github/gh-aw/pkg/linters/httprespbodyclose" | ||
| "github.com/github/gh-aw/pkg/linters/httpstatuscode" | ||
| "github.com/github/gh-aw/pkg/linters/jsonmarshalignoredeerror" |
| // GoodNoClose returns the response to the caller, which is responsible for closing — not flagged. | ||
| func GoodNoClose(client *http.Client, req *http.Request) (*http.Response, error) { | ||
| return client.Do(req) | ||
| } |
🧪 Test Quality Sentinel Report✅ Test Quality Score: 100/100 — Excellent
📊 Metrics (1 test)
Notes:
Verdict
Warning Firewall blocked 1 domainThe following domain was blocked by the firewall during workflow execution:
network:
allowed:
- defaults
- "awmgmcpg"See Network Configuration for more information.
|
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
🏗️ Design Decision Gate — ADR RequiredThis PR makes significant changes to core business logic (248 new lines in 📄 Draft ADR committed:
📋 What to do next
Once an ADR is linked in the PR body, this gate will re-run and verify the implementation matches the decision. ❓ Why ADRs Matter
ADRs create a searchable, permanent record of why the codebase looks the way it does. Future contributors (and your future self) will thank you. 📋 Michael Nygard ADR Format ReferenceAn ADR must contain these four sections to be considered complete:
All ADRs are stored in
|
Skills-Based Review 🧠Applied 📋 Key Themes & HighlightsKey Themes
Positive Highlights
@copilot please address the review comments above.
|
There was a problem hiding this comment.
🧠 Reviewed using Matt Pocock's skills by Matt Pocock Skills Reviewer · 76 AIC · ⌖ 12.8 AIC · ⊞ 6.6K
Comment /matt to run again
| // GoodNoClose returns the response to the caller, which is responsible for closing — not flagged. | ||
| func GoodNoClose(client *http.Client, req *http.Request) (*http.Response, error) { | ||
| return client.Do(req) | ||
| } |
There was a problem hiding this comment.
[/tdd] Missing test case for closure/FuncLit boundary — the implementation explicitly stops traversal at *ast.FuncLit (line 78 of the analyzer) but there is no fixture verifying that a resp.Body.Close() inside a closure does not trigger a diagnostic on the outer response variable.
💡 Suggested test fixture
The analogous fileclosenotdeferred testdata has an OpenInClosure case for exactly this boundary. Add a similar case:
// GoodCloseInsideClosure — resp.Body.Close() inside a closure should not be// attributed to the outer function's resp variable.funcGoodCloseInsideClosure(client*http.Client, req*http.Request) {
resp, err:=client.Do(req)
iferr!=nil {
return
}
gofunc() { resp.Body.Close() }()
}Without this test, a regression in the FuncLit guard would go undetected.
@copilot please address this.
| // GoodNoClose returns the response to the caller, which is responsible for closing — not flagged. | ||
| func GoodNoClose(client *http.Client, req *http.Request) (*http.Response, error) { | ||
| return client.Do(req) | ||
| } |
There was a problem hiding this comment.
[/tdd] Missing test case for variable shadowing — the analogous fileclosenotdeferred testdata has a ShadowedVar case, but there is none here. A shadowed resp variable (inner block re-declaring with :=) could interact with the types.Object keying in unexpected ways.
💡 Suggested test fixture
// GoodShadowedResp — inner resp shadows outer resp; both are deferred — not flagged.funcGoodShadowedResp(client*http.Client, req1, req2*http.Request) error {
resp, err:=client.Do(req1)
iferr!=nil {
returnerr
}
deferresp.Body.Close()
iftrue {
resp, err:=client.Do(req2) // shadows outer respiferr!=nil {
returnerr
}
deferresp.Body.Close()
_=resp
}
returnnil
}@copilot please address this.
| // GoodNoClose returns the response to the caller, which is responsible for closing — not flagged. | ||
| func GoodNoClose(client *http.Client, req *http.Request) (*http.Response, error) { | ||
| return client.Do(req) | ||
| } |
There was a problem hiding this comment.
[/tdd] Missing test case for response reassignment — the fileclosenotdeferred testdata has ReopenWithManualCloseThenDefer to cover the variable-reuse path in trackHTTPAssignment (lines 148–155 of the analyzer). No equivalent fixture is present here, leaving the "report prior unresolved violation before overwriting" logic untested.
💡 Suggested test fixture
// BadReopenManualCloseThenDefer — first response manually closed (violation),// second response deferred (ok). First should be flagged.funcBadReopenManualCloseThenDefer(client*http.Client, req1, req2*http.Request) error {
resp, err:=client.Do(req1) // want `HTTP response Body\.Close\(\) should be deferred.*`iferr!=nil {
returnerr
}
resp.Body.Close() // manual close — violation for first callresp, err=client.Do(req2)
iferr!=nil {
returnerr
}
deferresp.Body.Close()
returnnil
}@copilot please address this.
| // GoodNoClose returns the response to the caller, which is responsible for closing — not flagged. | ||
| func GoodNoClose(client *http.Client, req *http.Request) (*http.Response, error) { | ||
| return client.Do(req) | ||
| } |
There was a problem hiding this comment.
[/tdd] Missing nolint suppression test cases — the testdata has no fixture for //nolint:httprespbodyclose, despite the analyzer supporting suppression. Both code paths that check nolint.HasDirective (line 64 and line 149) are left untested.
💡 Suggested test fixtures
// SuppressedManualClose — nolint directive suppresses the violation — not flagged.funcSuppressedManualClose(client*http.Client, req*http.Request) ([]byte, error) {
resp, err:=client.Do(req) (nolint/redacted):httprespbodycloseiferr!=nil {
returnnil, err
}
data, _:=io.ReadAll(resp.Body)
resp.Body.Close()
returndata, nil
}Without this, a bug in the nolint wiring would only be caught by the broader integration test suite (if it covers this linter at all).
@copilot please address this.
| if !nolint.HasDirective(pass.Fset.PositionFor(state.assignPos, false), noLintLinesByFile) { | ||
| pass.Report(analysis.Diagnostic{ | ||
| Pos: state.assignPos, | ||
| Message: "HTTP response Body.Close() should be deferred immediately after receiving the response to prevent resource leaks", |
There was a problem hiding this comment.
[/tdd] Diagnostic reported at the call-site position, not at the non-deferred Body.Close() line — this may be confusing to users. The message says "should be deferred immediately after receiving the response" but the caret in the IDE/CLI points to the HTTP call, not the manual Body.Close().
💡 Discussion
Pointing to the assignment line is reasonable (it's where the resource is acquired), but the existing fileclosenotdeferred analyzer follows the same convention, so this is consistent within the codebase. Worth considering whether a secondary note pointing to the actual Body.Close() line (using analysis.Diagnostic.Related) would improve developer experience, similar to how the Go compiler uses "note" annotations. Not a blocking issue but a UX improvement opportunity.
@copilot please address this.
| if obj := bodyCloseReceiver(pass, deferStmt.Call); obj != nil { | ||
| if state, found := respVars[obj]; found { | ||
| state.hasDeferClose = true | ||
| } |
There was a problem hiding this comment.
[/codebase-design]defer seen before the response is tracked — potential false negative. The inspectNode function processes ast.AssignStmt (line 82) and ast.DeferStmt (line 92) in AST traversal order within the same ast.Inspect walk. For the happy path (resp, err := ... followed by defer resp.Body.Close()) this works correctly because assignments always appear before the defer in source order. However, consider: what happens if code does defer resp.Body.Close() before resp is assigned in the same scope? The respVars[obj] lookup at line 94 will not find the key, silently ignoring the defer. This is an unlikely anti-pattern but worth a comment explaining the ordering assumption.
💡 Suggested improvement
Add a brief inline comment at line 92 clarifying the traversal order assumption:
// defer statements are processed in source order; if a defer appears before// the assignment (unusual) the state lookup below will be a no-op, which is safe.ifdeferStmt, ok:=node.(*ast.DeferStmt); ok {@copilot please address this.
| if assign, ok := node.(*ast.AssignStmt); ok { | ||
| trackHTTPAssignment(pass, respVars, assign, noLintLinesByFile) | ||
| // Also check RHS for manual Body.Close() in assignments like: err := resp.Body.Close() |
There was a problem hiding this comment.
[/tdd] Missing test case for err := resp.Body.Close() — the comment on line 84 says the RHS-of-assignment path handles err := resp.Body.Close(), but the testdata has no fixture exercising it. This path is distinct from the exprStmt path (line 100) and a bug here would go undetected.
💡 Suggested test fixture
// BadManualCloseAssigned — resp.Body.Close() result captured in a variable — also flagged.funcBadManualCloseAssigned(client*http.Client, req*http.Request) ([]byte, error) {
resp, err:=client.Do(req) // want `HTTP response Body\.Close\(\) should be deferred.*`iferr!=nil {
returnnil, err
}
data, _:=io.ReadAll(resp.Body)
err=resp.Body.Close()
returndata, err
}@copilot please address this.
gh-aw-bot
commented
Jul 1, 2026
@copilot please run the
Run: https://github.com/github/gh-aw/actions/runs/28555444682
|
Co-authored-by: gh-aw-bot <259018956+gh-aw-bot@users.noreply.github.com>
Addressed in
I also attempted to update the PR body to add the ADR link, but this environment’s |
🎉 This pull request is included in a new release. Release: |
Summary
Adds a new custom Go static analysis linter,
httprespbodyclose, to the gh-aw linter suite. The analyzer enforces thatresp.Body.Close()calls on*net/http.Responsevariables are always made viadefer, preventing resource leaks caused by non-deferred closes when a function returns early.What changed
pkg/linters/httprespbodyclose/cmd/linters/main.gohttprespbodyclose.Analyzerin the multicheckerpkg/cli/mcp_inspect_mcp_scripts_server.goresp.Body.Close()inwaitForServerReadydocs/adr/42816-add-httprespbodyclose-linter.mdpkg/linters/httprespbodyclose/httprespbodyclose_test.goanalysistestharnesspkg/linters/httprespbodyclose/testdata/src/httprespbodyclose/httprespbodyclose.goAnalyzer design
FuncDecltraversal; stops atFuncLitboundaries (closures treated as separate scope)go/typesto confirm*net/http.Responsefromnet/httppackage — not syntactic string matchingrespVarsmap (types.Object → respVarState) records assignment position, defer close, and manual closeBody.Close()call siterespvariable is reassigned//nolint:httprespbodycloseon the assignment linefilecheck.IsTestFile, consistent with other suite analyzersTest case coverage
BadManualCloseBody.Close()GoodDeferClosedefer Body.Close()GoodNoCloseGoodCloseInsideClosureGoodShadowedRespBadReopenManualCloseThenDeferBadManualCloseAssignedBody.Close()on assignment RHSSuppressedOnAssignment//nolint:httprespbodycloseSuppressedAcrossReassignmentDesign decisions (ADR-42816)
Why not
timakin/bodyclose? The gh-aw suite uses a custom multichecker with shared internals (astutil,filecheck,nolint). Adopting a third-party linter would bypass this infrastructure and introduce an external dependency.Why not extend
fileclosenotdeferred?*http.Responsehas different semantics fromos.Open-family handles and requiresnet/httptype resolution. Keeping them separate preserves clarity and independent testability.Known limitations
*http.Responseacross function boundaries (cross-function close via helper → false negative)respfrom an outer scope (conservative: closure boundary terminates descent)Commits
6f523b20cfeat(linters): add httprespbodyclose linter1ba3c1350docs(adr): add draft ADR-42816 for httprespbodyclose linter9c6d53b32test(linters): cover httprespbodyclose edge cases and finalize ADRReferences:§28558128744