-
Notifications
You must be signed in to change notification settings - Fork 1.9k
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.
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
Improve PR review experience #9699
Comments
+1 for this, especially true for critical / long-running PRs. I haven't tested on the feasibility of this yet, but does it make sense to create a new GitHub issue (with relevant tags added) for each PR (similar to the auto backport PRs for example), wherein all of the GitHub actions can post the results etc.? The issue description could house the latest info from the most recent action runs. On the PR side, there could be a validation to resolve open comments on the linked issue before merging the PR. This would also keep the PR conversation relatively more clean with only the review comments displayed. 99% of the times, we are only looking for the results from the most recent runs. Its only when we want to maybe fetch a specific test seed or go back for another issue we observed in a previous build/action that we actually go looking for the previous build/action results. |
I like the simple proposal of removing green noise! Another idea is for gradle check to be deleting, or editing the previous gradle check comment, regardless of its status, so you only see the latest. This is something that danger does. |
Gradle checks take 30+ minutes to complete. It is quite common to push a commit to a PR, then quickly discover a new change is needed and push an updated commit, which results in two workflows running and both post a result comment to the PR. By deduping and canceling we will reduce load on the GitHub action runner (probably not on the Jenkins server unless the cancellation can propagate to Jenkins) and reduce noise on the PR. Relates opensearch-project#9699 Signed-off-by: Andrew Ross <[email protected]>
@BhumikaSaini-Amazon This is an interesting idea, but I'm inclined not to do this because the information that is posted in comments is already available elsewhere (in the list of github action workflows at the bottom of the PR), it just takes several clicks to get to the relevant information. The reason we post comments is to highlight the information and make it accessible with fewer clicks, I just think we've gone way too far with this. |
@dblock I think this is possible with the create-or-update workflow we're using (see example). The biggest challenge might be race conditions because it is rather important that only the latest run update the single comment. |
Instead of appending a new comment every time check compatibility runs this will instead update a single comment with the latest results. The history of previous runs is retained in the comment as GitHub allows you to view the edited revisions of any comment. This matches the behavior of the codecov workflow and should substantially reduce noise in PRs. Relates opensearch-project#9699 Signed-off-by: Andrew Ross <[email protected]>
Instead of appending a new comment every time check compatibility runs, this will instead update a single comment with the latest results. The history of previous runs is retained in the comment as GitHub allows you to view the edited revisions of any comment. This matches the behavior of the codecov workflow and should substantially reduce noise in PRs. Relates opensearch-project#9699 Signed-off-by: Andrew Ross <[email protected]>
Instead of appending a new comment every time check compatibility runs, this will instead update a single comment with the latest results. The history of previous runs is retained in the comment as GitHub allows you to view the edited revisions of any comment. This matches the behavior of the codecov workflow and should substantially reduce noise in PRs. Relates opensearch-project#9699 Signed-off-by: Andrew Ross <[email protected]>
Instead of appending a new comment every time check compatibility runs, this will instead update a single comment with the latest results. The history of previous runs is retained in the comment as GitHub allows you to view the edited revisions of any comment. This matches the behavior of the codecov workflow and should substantially reduce noise in PRs. Relates #9699 Signed-off-by: Andrew Ross <[email protected]>
Instead of appending a new comment every time check compatibility runs, this will instead update a single comment with the latest results. The history of previous runs is retained in the comment as GitHub allows you to view the edited revisions of any comment. This matches the behavior of the codecov workflow and should substantially reduce noise in PRs. Relates #9699 Signed-off-by: Andrew Ross <[email protected]> (cherry picked from commit c100c0c) Signed-off-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Instead of appending a new comment every time check compatibility runs, this will instead update a single comment with the latest results. The history of previous runs is retained in the comment as GitHub allows you to view the edited revisions of any comment. This matches the behavior of the codecov workflow and should substantially reduce noise in PRs. Relates #9699 (cherry picked from commit c100c0c) Signed-off-by: Andrew Ross <[email protected]> Signed-off-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Instead of appending a new comment every time check compatibility runs, this will instead update a single comment with the latest results. The history of previous runs is retained in the comment as GitHub allows you to view the edited revisions of any comment. This matches the behavior of the codecov workflow and should substantially reduce noise in PRs. Relates opensearch-project#9699 Signed-off-by: Andrew Ross <[email protected]> Signed-off-by: Kaushal Kumar <[email protected]>
This is the same change that was applied to the check comptability task by PR opensearch-project#9954 to update a single comment with the latest result of the gradle check workflow. Relates opensearch-project#9699 Signed-off-by: Andrew Ross <[email protected]>
This is the same change that was applied to the check comptability task by PR opensearch-project#9954 to update a single comment with the latest result of the gradle check workflow. Relates opensearch-project#9699 Signed-off-by: Andrew Ross <[email protected]>
This is the same change that was applied to the check comptability task by PR opensearch-project#9954 to update a single comment with the latest result of the gradle check workflow. Relates opensearch-project#9699 Signed-off-by: Andrew Ross <[email protected]>
The summary here is that the "check compatibility" workflow has been updated to update a single comment, bringing it in line with the codecov behavior, which should substantially reduce noise on the PR comment feed. However, multiple contributors mentioned that they prefer that "gradle check" create a new comment for every run so that it generates a notification that the PR needs attention (either success or failure). There were also opinions that it is desirable to not cancel concurrent gradle check runs because the additional runs can help find flaky tests. |
Instead of appending a new comment every time check compatibility runs, this will instead update a single comment with the latest results. The history of previous runs is retained in the comment as GitHub allows you to view the edited revisions of any comment. This matches the behavior of the codecov workflow and should substantially reduce noise in PRs. Relates opensearch-project#9699 Signed-off-by: Andrew Ross <[email protected]>
Instead of appending a new comment every time check compatibility runs, this will instead update a single comment with the latest results. The history of previous runs is retained in the comment as GitHub allows you to view the edited revisions of any comment. This matches the behavior of the codecov workflow and should substantially reduce noise in PRs. Relates opensearch-project#9699 Signed-off-by: Andrew Ross <[email protected]> Signed-off-by: Ivan Brusic <[email protected]>
Continuing in the spirit of opensearch-project#9699, this makes the gradle check comment less verbose while containing all the same information. The goal is to reduce visual noise on the PR feed and maybe allow for more comments before GitHub starts hiding older comments behind the "load more" link. This is super subjective so happy to hear any opinions if the previous format is preferred. Signed-off-by: Andrew Ross <[email protected]>
Continuing in the spirit of #9699, this makes the gradle check comment less verbose while containing all the same information. The goal is to reduce visual noise on the PR feed and maybe allow for more comments before GitHub starts hiding older comments behind the "load more" link. This is super subjective so happy to hear any opinions if the previous format is preferred. Signed-off-by: Andrew Ross <[email protected]>
Continuing in the spirit of opensearch-project#9699, this makes the gradle check comment less verbose while containing all the same information. The goal is to reduce visual noise on the PR feed and maybe allow for more comments before GitHub starts hiding older comments behind the "load more" link. This is super subjective so happy to hear any opinions if the previous format is preferred. Signed-off-by: Andrew Ross <[email protected]> (cherry picked from commit dcb64d8)
Continuing in the spirit of #9699, this makes the gradle check comment less verbose while containing all the same information. The goal is to reduce visual noise on the PR feed and maybe allow for more comments before GitHub starts hiding older comments behind the "load more" link. This is super subjective so happy to hear any opinions if the previous format is preferred. Signed-off-by: Andrew Ross <[email protected]> (cherry picked from commit dcb64d8)
Continuing in the spirit of opensearch-project#9699, this makes the gradle check comment less verbose while containing all the same information. The goal is to reduce visual noise on the PR feed and maybe allow for more comments before GitHub starts hiding older comments behind the "load more" link. This is super subjective so happy to hear any opinions if the previous format is preferred. Signed-off-by: Andrew Ross <[email protected]>
Continuing in the spirit of opensearch-project#9699, this makes the gradle check comment less verbose while containing all the same information. The goal is to reduce visual noise on the PR feed and maybe allow for more comments before GitHub starts hiding older comments behind the "load more" link. This is super subjective so happy to hear any opinions if the previous format is preferred. Signed-off-by: Andrew Ross <[email protected]>
Instead of appending a new comment every time check compatibility runs, this will instead update a single comment with the latest results. The history of previous runs is retained in the comment as GitHub allows you to view the edited revisions of any comment. This matches the behavior of the codecov workflow and should substantially reduce noise in PRs. Relates opensearch-project#9699 Signed-off-by: Andrew Ross <[email protected]> Signed-off-by: Shivansh Arora <[email protected]>
Continuing in the spirit of opensearch-project#9699, this makes the gradle check comment less verbose while containing all the same information. The goal is to reduce visual noise on the PR feed and maybe allow for more comments before GitHub starts hiding older comments behind the "load more" link. This is super subjective so happy to hear any opinions if the previous format is preferred. Signed-off-by: Andrew Ross <[email protected]> Signed-off-by: Shivansh Arora <[email protected]>
The most important information on the front page of a PR is the author's description of the change and the reviewers' comments. However, we have made liberal use of the "create-or-update" GitHub action for our various workflows to post information directly on the PR. This is all well intentioned and there is important information being posted, but I believe we've probably gone to far and the single-to-noise ratio here has dropped, and too often comments are obscured by the dreaded "load more" button. Also my scroll wheel finger is getting quite the workout :) Here are some specific thoughts I have to improve things:
Any other ideas?
The text was updated successfully, but these errors were encountered: