Skip to content

fix(ci): run go mod tidy after storage-e2e replace to survive dependency graph drift #56

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
fastrapier wants to merge 1 commit into
base: main
Choose a base branch
from
Draft

fix(ci): run go mod tidy after storage-e2e replace to survive dependency graph drift #56

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 14 additions & 6 deletions .github/scripts/e2e-run-tests.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,11 +7,15 @@
# up before editing and restored via a trap on EXIT, so the working tree is not
# left mutated on self-hosted runners (defensive even though checkout uses
# clean: true; a caller may still override it to clean: false). After
# applying the replace we run `go mod download` (not `go mod tidy`): download is
# deterministic and fast, whereas tidy re-resolves the whole graph and can stall
# fetching modules on the runner. The consumer is responsible for keeping its
# go.mod/go.sum in sync with the storage-e2e revision it pins (so the replaced
# module's deps are already covered).
# applying the replace we run `go mod tidy` (not `go mod download`): the
# directory replace points at storage-e2e@main (and PR merge commits may carry
# further local replaces such as ../api), so the effective dependency graph can
# differ from what the consumer's go.mod/go.sum were resolved against. download
# does not re-resolve the graph, and MVS may then pick module versions missing
# from the consumer's go.sum, failing with "missing go.sum entry for module
# providing package ...". tidy re-resolves go.mod/go.sum against the actual
# graph; mutating them is safe because both are backed up and restored by the
# EXIT trap below.
#
# Inputs (env):
# E2E_MODULE_PATH path to the Go module under test (default ".")
Expand Down Expand Up @@ -55,13 +59,17 @@ if [ "$module_name" != "github.com/deckhouse/storage-e2e" ]; then
if [ -n "$gosum_backup" ]; then
cp "$gosum_backup" go.sum
rm -f "$gosum_backup"
else
# No go.sum existed before; `go mod tidy` may have created one. Remove it
# so the working tree is left unchanged.
rm -f go.sum
fi
return $rc
}
trap restore_gomod EXIT

"$go_bin" mod edit -replace="github.com/deckhouse/storage-e2e=${E2E_STORAGE_E2E_DIR}"
"$go_bin" mod download
"$go_bin" mod tidy
fi

echo "Ginkgo label filter: ${label_filter}"
Expand Down
50 changes: 34 additions & 16 deletions .github/scripts/tests/test-run-tests.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,9 +1,10 @@
#!/usr/bin/env bash
# Tests for e2e-run-tests.sh using a stub `go` (via GO_BIN) that records its
# invocations. Verifies: replace is applied for a consumer module and skipped
# for storage-e2e itself; `go mod download` (not `go mod tidy`) is used; the
# consumer module's go.mod is restored after the run (trap on EXIT); the Ginkgo
# label filter is passed; a log file is written.
# for storage-e2e itself; `go mod tidy` (not `go mod download`) is used; the
# consumer module's go.mod/go.sum are restored after the run (trap on EXIT),
# including removal of a go.sum created by tidy when none existed before; the
# Ginkgo label filter is passed; a log file is written.
set -euo pipefail

SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
Expand All @@ -13,41 +14,58 @@ pass() { echo "PASS: $1"; }
err() { echo "FAIL: $1"; fail=1; }

make_stub() { # stub_path log_path module_name
# The stub records every invocation. For `mod edit`/`mod download` it appends
# a marker line to go.mod (cwd is the module dir) to simulate the real go
# mutating go.mod, so the test can verify the trap restores the original file.
# The stub records every invocation. For `mod edit`/`mod tidy` it mutates
# go.mod and go.sum (cwd is the module dir) to simulate the real go
# re-resolving the graph, so the test can verify the trap restores the
# original files (and removes a go.sum created by tidy when none existed).
cat >"$1" <<EOF
#!/usr/bin/env bash
echo "\$*" >> "$2"
case "\$1 \$2" in
"list -m") echo "$3" ;;
"mod edit"*) echo "// e2e-stub-edit-mutation" >> go.mod ;;
"mod download"*) echo "// e2e-stub-download-mutation" >> go.mod ;;
"test"*) exit 0 ;;
*) exit 0 ;;
"list -m") echo "$3" ;;
"mod edit"*) echo "// e2e-stub-edit-mutation" >> go.mod ;;
"mod tidy"*) echo "// e2e-stub-tidy-mutation" >> go.mod
echo "example.com/dep v1.0.0 h1:stub" >> go.sum ;;
"test"*) exit 0 ;;
*) exit 0 ;;
esac
EOF
chmod +x "$1"
}

# Case 1: consumer module -> replace applied, filter passed, log written,
# `go mod download` used (not tidy), and go.mod restored afterwards.
work="$(mktemp -d)"; mod="$(mktemp -d)"; stub_log="$(mktemp)"; stub="$(mktemp)"; gomod_orig="$(mktemp)"
# `go mod tidy` used (not download), and go.mod/go.sum restored afterwards.
work="$(mktemp -d)"; mod="$(mktemp -d)"; stub_log="$(mktemp)"; stub="$(mktemp)"; gomod_orig="$(mktemp)"; gosum_orig="$(mktemp)"
printf 'module example.com/consumer\n' >"${mod}/go.mod"
printf 'example.com/old v0.1.0 h1:orig\n' >"${mod}/go.sum"
cp "${mod}/go.mod" "$gomod_orig"
cp "${mod}/go.sum" "$gosum_orig"
make_stub "$stub" "$stub_log" "example.com/consumer"
GITHUB_WORKSPACE="$work" GO_BIN="$stub" \
E2E_MODULE_PATH="$mod" E2E_TEST_PACKAGE="./tests/" \
E2E_GINKGO_LABEL_FILTER="stress-test || integration" \
E2E_STORAGE_E2E_DIR="/tmp/storage-e2e" \
bash "$SCRIPT" >/dev/null || true
grep -q 'mod edit -replace=github.com/deckhouse/storage-e2e=/tmp/storage-e2e' "$stub_log" && pass "replace applied" || err "replace missing"
grep -q 'mod download' "$stub_log" && pass "mod download used" || err "mod download missing"
if grep -q 'mod tidy' "$stub_log"; then err "mod tidy should not be used"; else pass "mod tidy not used"; fi
grep -q 'mod tidy' "$stub_log" && pass "mod tidy used" || err "mod tidy missing"
if grep -q 'mod download' "$stub_log"; then err "mod download should not be used"; else pass "mod download not used"; fi
grep -q 'ginkgo.label-filter=stress-test || integration' "$stub_log" && pass "label filter passed" || err "label filter missing"
[ -f "${work}/e2e-test-output.log" ] && pass "test log written" || err "test log missing"
if cmp -s "${mod}/go.mod" "$gomod_orig"; then pass "go.mod restored after run"; else err "go.mod not restored after run"; fi
rm -rf "$work" "$mod" "$stub_log" "$stub" "$gomod_orig"
if cmp -s "${mod}/go.sum" "$gosum_orig"; then pass "go.sum restored after run"; else err "go.sum not restored after run"; fi
rm -rf "$work" "$mod" "$stub_log" "$stub" "$gomod_orig" "$gosum_orig"

# Case 1b: consumer module without go.sum -> a go.sum created by tidy is
# removed on exit so the working tree is left unchanged.
work="$(mktemp -d)"; mod="$(mktemp -d)"; stub_log="$(mktemp)"; stub="$(mktemp)"
printf 'module example.com/consumer\n' >"${mod}/go.mod"
make_stub "$stub" "$stub_log" "example.com/consumer"
GITHUB_WORKSPACE="$work" GO_BIN="$stub" \
E2E_MODULE_PATH="$mod" E2E_TEST_PACKAGE="./tests/" \
E2E_STORAGE_E2E_DIR="/tmp/storage-e2e" \
bash "$SCRIPT" >/dev/null || true
if [ -f "${mod}/go.sum" ]; then err "go.sum created by tidy should be removed"; else pass "go.sum created by tidy removed"; fi
rm -rf "$work" "$mod" "$stub_log" "$stub"

# Case 2: storage-e2e itself -> replace skipped.
work="$(mktemp -d)"; mod="$(mktemp -d)"; stub_log="$(mktemp)"; stub="$(mktemp)"
Expand Down
Loading