From 91000b3449698ecce9d256fc2fec4aa5ea31c13b Mon Sep 17 00:00:00 2001 From: tannevaled Date: Wed, 11 Nov 2026 16:09:09 +0100 Subject: [PATCH 1/2] ci: install poppler through aptneed, so a mirror outage stops blocking releases MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit ⛔⛔ A THIRD PARTY'S DNS HELD EVERY RELEASE HOSTAGE. On 2026-11-11 this step sudo apt-get update && sudo apt-get install -y --no-install-recommends poppler-utils failed FOUR times in half an hour on one merge commit, every time with `Temporary failure resolving 'azure.archive.ubuntu.com'`, while the pull request it came from had been green on this same workflow minutes earlier — and then started failing pull requests too. ghrelease will not tag while main is red, correctly, so nothing about this repository was wrong, nothing could be released, and the step had no way to say which of the two it was (#90). aptneed (go-fleettools/fleettools) rewrites /etc/apt/apt-mirrors.txt to the canonical archive after the first mirror failure — that one file is what every source on a runner resolves through, so naming archive.ubuntu.com in a SOURCE would change nothing — and it refuses to retry a package apt cannot locate, which is a name and not weather. ⛔ --need pdftoppm, because apt saying yes is not the binary being there. The two render tests already t.Fatal when CI is set and pdftoppm is missing (place_test.go:410, poster_render_test.go:114), which is the gate that keeps the foreign judge from being silently skipped; --need is the same question asked one step earlier, where it can still be answered. actionlint clean. Closes #90. Co-Authored-By: Claude Opus 5 --- .github/workflows/ci.yml | 22 +++++++++++++++++++++- 1 file changed, 21 insertions(+), 1 deletion(-) diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml index 90a37e7..a74912c 100644 --- a/.github/workflows/ci.yml +++ b/.github/workflows/ci.yml @@ -35,8 +35,28 @@ jobs: check-latest: true cache: true + # ⛔⛔ THIS STEP USED TO BE `sudo apt-get update && sudo apt-get install …` + # AND A THIRD PARTY'S DNS HELD EVERY RELEASE HOSTAGE. On 2026-11-11 it + # failed four times in half an hour on one merge commit, every time with + # `Temporary failure resolving 'azure.archive.ubuntu.com'`, while the + # pull request it came from had been green on this same workflow minutes + # earlier. `ghrelease` will not tag while main is red — correctly — so + # nothing about this repository was wrong and nothing could be released, + # and the step had no way to say which of the two it was (#90). + # + # `aptneed` rewrites /etc/apt/apt-mirrors.txt to the canonical archive + # after the first mirror failure — that one file is what every source on + # a runner resolves through, so naming archive.ubuntu.com in a source + # would change nothing — and it refuses to retry a package apt cannot + # locate, which is a name and not weather. + # + # ⛔ --need pdftoppm, because apt saying yes is not the binary being + # there. The two render tests `t.Fatal` when CI is set and pdftoppm is + # missing, so a package renamed upstream would otherwise reach them. - name: Install poppler, the foreign judge the render tests print by - run: sudo apt-get update && sudo apt-get install -y --no-install-recommends poppler-utils + run: | + GOBIN=/usr/local/bin go install github.com/go-fleettools/fleettools/cmd/aptneed@latest + sudo aptneed --need pdftoppm poppler-utils - name: go vet run: go vet ./... From 889bde3e783b58a1c1cc6e718436660f86cde695 Mon Sep 17 00:00:00 2001 From: tannevaled Date: Fri, 13 Nov 2026 18:36:13 +0100 Subject: [PATCH 2/2] ci: Go 1.27.4, which is what installing aptneed actually requires MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit ⛔ THE FIRST PUSH OF THIS BRANCH WENT RED, AND NOT FOR THE REASON I HAD GUARDED AGAINST. I held it back until go-fleettools/fleettools v0.111.0 was released AND `go list -m …@latest` resolved to it, because `@latest` is a tag and the module proxy serves it from its own cache. That part worked: the runner logged `go: downloading …/fleettools v0.111.0`. What failed is one line further down: go: …/cmd/aptneed@latest: …/fleettools@v0.111.0 requires go >= 1.27.4 (running go 1.27.3; GOTOOLCHAIN=local) fleettools v0.111.0 declares `go >= 1.27.4` and this workflow pinned 1.27.3. ⭐ AND GOTOOLCHAIN=local IS WHY THIS IS A CLEAN FAILURE RATHER THAN A SILENT SWAP. Without it Go would have downloaded 1.27.4 on its own and the job would have measured a toolchain nobody chose — the thing this repository sets that variable to prevent. The refusal names both versions and the remedy is one line, which is what a pinned toolchain is for. go.mod still says `go 1.27.2` and is left alone: that is the language version this module needs, not the toolchain the runner must have, and nothing here requires it to move. actionlint clean. Co-Authored-By: Claude Opus 5 --- .github/workflows/ci.yml | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml index a74912c..30b8263 100644 --- a/.github/workflows/ci.yml +++ b/.github/workflows/ci.yml @@ -29,9 +29,14 @@ jobs: steps: - uses: actions/checkout@v7 + # ⛔ 1.27.4 BECAUSE THE STEP BELOW NEEDS IT, and GOTOOLCHAIN=local means + # the runner will not quietly fetch a newer one — which is the setting + # that made this a visible failure instead of a silent toolchain swap. + # fleettools v0.111.0 declares `go >= 1.27.4`, so 1.27.3 refused to + # install aptneed at all: `requires go >= 1.27.4 (running go 1.27.3)`. - uses: actions/setup-go@v7 with: - go-version: '1.27.3' + go-version: '1.27.4' check-latest: true cache: true