Skip to content

fix(actions): Lopu repairs failed PR checks #577

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

Merged
lopugit merged 1 commit into from
Sep 3, 2026
Merged

fix(actions): Lopu repairs failed PR checks #577

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
26 changes: 26 additions & 0 deletions .github/scripts/resolve-pr-conflicts-routing-contract.mjs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -516,6 +516,32 @@ function assertWorkflowSource() {
/if grep -Eq 'HTTP \(408\|429\|500\|502\|503\|504\)\(\[\^0-9\]\|\$\)' "\$errors"; then/u,
"no gh_read_retry copy classifies transience by HTTP status alone",
);
// GitHub answers a diff it could not finish generating with HTTP 422 and
// "Sorry, this diff is taking too long to generate." That is a server-side
// generation timeout on an idempotent read, not a client error: the
// paginated `pulls/291/files` read that took the repository-wide scan down
// in run 33565995120 replayed cleanly minutes later, all 29 pages. Pin the
// declaration and the branch separately, exactly like the transport
// predicate, because a copy that keeps one without the other is silently
// back to the outage shape.
assert.equal(
source.match(/^\s*slow_diff='diff is taking too long to generate'$/gmu)?.length,
readRetryCopies,
"every gh_read_retry copy declares GitHub's diff-generation timeout",
);
assert.equal(
source.match(/&& grep -Fq "\$slow_diff" "\$errors"/gu)?.length,
readRetryCopies,
"every gh_read_retry copy branches on the diff-generation timeout, not just declares it",
);
// The status alone must never become retryable: an invalid ref, "No commits
// between", and a failed validation are all 422 and all have to stay fatal
// on the first attempt. The message is only honoured paired with its status.
assert.equal(
source.match(/grep -Eq 'HTTP 422\(\[\^0-9\]\|\$\)' "\$errors"/gu)?.length,
readRetryCopies,
"every gh_read_retry copy scopes that retry to a 422 carrying that exact message",
);
assert.match(
source,
/if ! gh_read_retry graphql --paginate --slurp[\s\S]*Could not read the open PR inventory from GitHub[\s\S]*successful but malformed PR inventory response/u,
Expand Down
38 changes: 31 additions & 7 deletions .github/workflows/resolve-pr-conflicts.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1407,15 +1407,27 @@ jobs:
# Match that decode message too: only a body that stopped early
# produces it, while a genuinely malformed payload (an HTML error
# page, an auth/permission JSON error) fails with a different message
# and still surfaces immediately. Keep every notice on stderr so
# and still surfaces immediately. A diff GitHub could not finish
# generating is the same class one layer further up: run 33565995120
# lost `pulls/291/files` (2896 files) to HTTP 422 "Sorry, this diff
# is taking too long to generate", which is a server-side generation
# timeout on an idempotent read, not a client validation error — the
# identical paginated read replayed cleanly minutes later, all 29
# pages. 422 as a status is NOT retryable (an invalid ref, "No
# commits between", a failed validation all use it and must still
# fail on the first attempt), so the status and that exact message
# are matched together. gh prints both on one line but in the
# opposite order, which is why this stays two tests, not one regex.
# Keep every notice on stderr so
# command-substitution JSON stays clean, and surface
# auth/schema/permission failures
# immediately. Mutating API calls deliberately do not use this helper
# because a blind retry could duplicate a side effect; every read
# routed here is idempotent, so replaying a torn-down request is safe.
gh_read_retry() {
local attempt output errors delay transport
local attempt output errors delay transport slow_diff
transport='stream error|http2: server sent GOAWAY|connection reset by peer|unexpected EOF|unexpected end of JSON input|TLS handshake timeout|i/o timeout|server closed idle connection|client connection force closed'
slow_diff='diff is taking too long to generate'
output="$RUNNER_TEMP/gh-read-${BASHPID:-$$}.out"
errors="$RUNNER_TEMP/gh-read-${BASHPID:-$$}.err"
for attempt in 1 2 3 4; do
Expand All @@ -1427,7 +1439,9 @@ jobs:
return 0
fi
if grep -Eq 'HTTP (408|429|500|502|503|504)([^0-9]|$)' "$errors" \
|| grep -Eq "$transport" "$errors"; then
|| grep -Eq "$transport" "$errors" \
|| { grep -Eq 'HTTP 422([^0-9]|$)' "$errors" \
&& grep -Fq "$slow_diff" "$errors"; }; then
if [ "$attempt" -lt 4 ]; then
delay="$((1 << attempt))"
echo "::notice::GitHub API read hit a transient failure (attempt $attempt/4); retrying in ${delay}s." >&2
Expand Down Expand Up @@ -3089,8 +3103,9 @@ jobs:
# Retry only read-only metadata commands and keep diagnostics off
# stdout so command-substitution JSON remains parseable.
gh_read_retry() {
local attempt delay output errors transport
local attempt delay output errors transport slow_diff
transport='stream error|http2: server sent GOAWAY|connection reset by peer|unexpected EOF|unexpected end of JSON input|TLS handshake timeout|i/o timeout|server closed idle connection|client connection force closed'
slow_diff='diff is taking too long to generate'
output="$RUNNER_TEMP/lopu-review-detect-gh-${BASHPID:-$$}.out"
errors="$RUNNER_TEMP/lopu-review-detect-gh-${BASHPID:-$$}.err"
for attempt in 1 2 3 4; do
Expand All @@ -3103,8 +3118,12 @@ jobs:
fi
# A transport reset never carries an HTTP status line; it is the
# same retryable edge failure as a 502/504 for an idempotent read.
# GitHub's diff-generation timeout does carry one, but 422 as a
# status is not retryable, so it is paired with its exact message.
if { grep -Eq 'HTTP (408|429|500|502|503|504)([^0-9]|$)' "$errors" \
|| grep -Eq "$transport" "$errors"; } \
|| grep -Eq "$transport" "$errors" \
|| { grep -Eq 'HTTP 422([^0-9]|$)' "$errors" \
&& grep -Fq "$slow_diff" "$errors"; }; } \
&& [ "$attempt" -lt 4 ]; then
delay="$((1 << attempt))"
echo "::notice::GitHub metadata read hit a transient failure (attempt $attempt/4); retrying in ${delay}s." >&2
Expand Down Expand Up @@ -4380,8 +4399,9 @@ jobs:
set -euo pipefail

gh_read_retry() {
local attempt delay output errors transport
local attempt delay output errors transport slow_diff
transport='stream error|http2: server sent GOAWAY|connection reset by peer|unexpected EOF|unexpected end of JSON input|TLS handshake timeout|i/o timeout|server closed idle connection|client connection force closed'
slow_diff='diff is taking too long to generate'
output="$RUNNER_TEMP/lopu-review-prepare-gh-${BASHPID:-$$}.out"
errors="$RUNNER_TEMP/lopu-review-prepare-gh-${BASHPID:-$$}.err"
for attempt in 1 2 3 4; do
Expand All @@ -4394,8 +4414,12 @@ jobs:
fi
# A transport reset never carries an HTTP status line; it is the
# same retryable edge failure as a 502/504 for an idempotent read.
# GitHub's diff-generation timeout does carry one, but 422 as a
# status is not retryable, so it is paired with its exact message.
if { grep -Eq 'HTTP (408|429|500|502|503|504)([^0-9]|$)' "$errors" \
|| grep -Eq "$transport" "$errors"; } \
|| grep -Eq "$transport" "$errors" \
|| { grep -Eq 'HTTP 422([^0-9]|$)' "$errors" \
&& grep -Fq "$slow_diff" "$errors"; }; } \
&& [ "$attempt" -lt 4 ]; then
delay="$((1 << attempt))"
echo "::notice::GitHub metadata read hit a transient failure (attempt $attempt/4); retrying in ${delay}s." >&2
Expand Down
Loading