Conversation
| recording=false | ||
|
|
||
| stop_rec() { [ "$recording" = true ] && "${cmd[@]}" jcmd 1 JFR.stop name="$profile"; } | ||
| trap stop_rec EXIT |
There was a problem hiding this comment.
[P1] Make this cleanup signal-safe and success-safe.
Interrupting the read with Ctrl-C terminates the script without running this EXIT trap, leaving the JFR recording active. A failed copy also leaves the remote /tmp artifact. On the normal success path, recording=false makes stop_rec return status 1, so the EXIT trap turns an otherwise successful run into a failure.
Smallest fix: use one idempotent cleanup function that stops an active recording, always removes the remote file, and returns 0; install it for EXIT, INT, TERM, and HUP.
| "${cmd[@]}" jcmd 1 VM.flags | ||
| "${cmd[@]}" jcmd 1 GC.heap_info | ||
| "${cmd[@]}" jcmd 1 GC.heap_dump "$tmpfile" | ||
| "${cmd[@]}" cat "$tmpfile" > "$dump" |
There was a problem hiding this comment.
[P1] Create diagnostic artifacts with private permissions.
With the normal 022 umask, this redirection creates a mode-0644 heap dump. Heap dumps can contain credentials and live application data; profile-jfr.sh:21 has the same issue for JFR data and recorded environment/system properties.
Smallest fix: set umask 077 before creating either output and ensure an existing destination is forced to mode 0600.
| --json nameWithOwner --jq '.[].nameWithOwner') | ||
|
|
||
| for repo in $repos; do | ||
| gh api "repos/$repo/branches/$branch" --silent >/dev/null 2>&1 || continue |
There was a problem hiding this comment.
[P1] Do not treat every API failure as “branch absent”.
Authentication failures, rate limits, API outages, and network errors all take this || continue path with stderr discarded. If every lookup fails, the batch prints nothing and exits successfully without creating any PRs.
Smallest fix: continue only for an explicit 404; surface other failures and return a non-zero status or a failure summary.
| ```bash | ||
| <toolbox>/mvn/build-dependency-tree.py ~/workspace > dependency.tree | ||
| ``` | ||
|
|
There was a problem hiding this comment.
[P1] Restore these workflows or provide their replacements in a dedicated retirement change.
This PR removes the documented prepare-release and major-upgrade workflows and deletes their scripts, along with the repository-bootstrap and deployment helpers, without migration guidance. The root README still advertises repository cloning, deployment helpers, and Maven release workflows, so users updating to this branch receive missing-command and stale-documentation failures unrelated to the profiling feature.
Smallest fix: restore these removals in this PR. If retirement is intentional, split it out and update the surviving contract with concrete replacement commands.
|
|
||
| for repo in $repos; do | ||
| gh api "repos/$repo/branches/$branch" --silent >/dev/null 2>&1 || continue | ||
| pr=$(gh pr list --repo "$repo" --head "$branch" --state open \ |
There was a problem hiding this comment.
[P2] Include the requested base branch in existing-PR detection.
The script forwards arbitrary gh pr create options, including --base, but this lookup filters only by head. If the same head already has an open PR to a different base, a request such as --base release is incorrectly skipped.
Smallest fix: determine the effective base and include it in gh pr list, or explicitly reject --base instead of advertising arbitrary create options.
2 participants
No description provided.