Skip to content

go 1.27.4 for GO-2026-6859, and the CI pin reads go.mod - #8

Merged
tannevaled merged 1 commit into
mainfrom
go1274
Nov 15, 2026
Merged

tannevaled merged 1 commit into
mainfrom
go1274

Conversation

@tannevaled

@tannevaled tannevaled commented Nov 15, 2026 •

Copy link
Copy Markdown
Contributor

Two halves, both required: the published floor, and the CI pin that would otherwise refuse it.

1. go 1.27.3 → go 1.27.4

GO-2026-6859, published 2026-11-11, CVE-2026-97054, "HTTP/1 server crash due to MaxBytesReader header race in net/http", standard library, fixed in 1.26.11 and 1.27.4 — read at vuln.go.dev rather than quoted from notes.

The go directive is the only one of the three Go pins a consumer reads: the CI pin builds our artefacts, the toolchain stamped in a binary describes one executable, and this line is what go install honours under GOTOOLCHAIN=local.

Measured: modfloor --floor 1.27.4 read the directive this module published at v0.3.1 as go 1.27.3, and reported it PUBLISHED — so a consumer installing it that way built against an unpatched standard library.

⚠ Reachability is govulncheck's question, not this line's. And unlike the eight modules raised earlier today, this repository is not in the fleet's network-facing list — the floor is raised because a consumer is entitled to build against the patched library, not on an exposure claim.

2. ⛔ The pin was hidden behind an env var

The workflow held GO_VERSION: "1.27.3" and its only setup-go step read ${{ env.GO_VERSION }} — a hard-coded pin one indirection away. setup-go sets GOTOOLCHAIN=local, so a pin under the directive does not build an older toolchain quietly; it refuses:

go: go.mod requires go >= 1.27.4 (running go 1.27.3; GOTOOLCHAIN=local)

Measured the same day on five of six go-filesystems repositories: every job red at once for a one-line change to go.mod. So both move together here.

⚠ And the indirection hid it from the auditor. gofloor reports a hard-coded go-version as a finding and classifies ${{ … }} merely as "expression", so this repository read as clean while its pin sat two patch releases behind the floor the fleet was moving to. Three of the five Go repositories in this organisation have that shape — browserproxy (fixed today), engine (current, invisible at the next raise), and this one. Being filed against gofloor separately.

The step now reads go-version-file: go.mod, so the toolchain is the declared floor by construction and the next raise needs no edit here.

⚠ One note about method, not about this change

This clone was four commits stale — 1585 insertions, cmd/webtest/ and driver/ added — and an edit made before fetching would have produced a pull request reverting all of it. Fetched and fast-forwarded to e5b0f39 before touching anything.

Measured

actionlint clean, gofmt -l . silent, go build ./... and go test ./... pass. The only go-version text left in the workflow is this change's own explanation.

Release level: MINOR

A raised floor genuinely breaks a consumer pinned below it with GOTOOLCHAIN=local: their build stops rather than quietly producing a vulnerable binary. The release is the fix, not the merge — a consumer reads the published go.mod.

🤖 Generated with Claude Code

… env var

GO-2026-6859, published 2026-11-11, CVE-2026-97054: "HTTP/1 server crash due to
MaxBytesReader header race in net/http", standard library, fixed in 1.26.11 and
1.27.4. Read at https://vuln.go.dev/ID/GO-2026-6859.json rather than quoted
from notes.

The `go` directive is the only one of the three Go pins a CONSUMER reads, and
`modfloor --floor 1.27.4` read the directive this module PUBLISHED at v0.3.1 as
`go 1.27.3`, reported PUBLISHED: a consumer installing it under
GOTOOLCHAIN=local built against an unpatched standard library.

⚠ Reachability is govulncheck's question, not this line's, and this repository
is NOT in the fleet's network-facing list. The floor is raised because a
consumer is entitled to build against the patched library.

⛔ AND THE CI PIN HAD TO MOVE IN THE SAME CHANGE. The workflow held
`GO_VERSION: "1.27.3"` and its only setup-go step read `${{ env.GO_VERSION }}`
— a hard-coded pin one indirection away. `setup-go` sets GOTOOLCHAIN=local, so
a pin under the directive does not build an older toolchain quietly, it REFUSES
with `go.mod requires go >= 1.27.4`. Measured the same day on five of six
go-filesystems repositories: every job red at once for a one-line change.

⚠ AND THE INDIRECTION HID IT FROM THE AUDITOR. `gofloor` reports a hard-coded
`go-version` as a finding and classifies `${{ … }}` merely as "expression", so
this repository read as clean while its pin sat two patch releases behind the
floor the fleet was moving to. Three of the five Go repositories in this
organisation have that shape — browserproxy (fixed today), engine (current but
invisible at the next raise), and this one. Being filed against gofloor
separately.

The step now reads `go-version-file: go.mod`, so the toolchain is the declared
floor by construction.

⚠ Worth recording about the method rather than the change: this clone was FOUR
COMMITS STALE (1585 insertions, cmd/webtest and driver/ added), and an edit made
before fetching would have produced a pull request reverting all of it. Fetched
and fast-forwarded to e5b0f39 before touching anything.

actionlint clean, gofmt silent, go build and go test ./... pass, and the only
`go-version` text left in the workflow is this change's own explanation.

MINOR, not patch: a raised floor is a genuine break for a consumer pinned below
it with GOTOOLCHAIN=local — their build stops instead of quietly producing a
vulnerable binary.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@tannevaled
tannevaled merged commit bcf28ba into main Nov 15, 2026
1 check passed
@tannevaled
tannevaled deleted the go1274 branch November 15, 2026 15:46
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.

1 participant