Handle closed pull requests and empty pull requests with no commit (#6108)

Fixes #6106. The issue is
caused by an closed empty pull request with no commit
pytorch/pytorch#143615.

* Handle a PR with no commit in `fetchPR`. This function is used by
Dr.CI to get all the commits in a pull request.
* Make sure that  Dr.CI only queries for open pull requests.
* Fix the broken HUD PR page https://hud.pytorch.org/pr/143615

### Testing
Both

```
curl --request POST \
    --url 'http://localhost:3000/api/drci/drci' \
    --header 'Authorization: <REDACTED>' \
    --data 'repo=pytorch'
```

and

```
curl --request POST \
    --url 'http://localhost:3000/api/drci/drci?prNumber=143615' \
    --header 'Authorization: <REDACTED>' \
    --data 'repo=pytorch'
```

are working ok now.


https://torchci-git-fork-huydhn-fix-drci-issue-6106-fbopensource.vercel.app/pr/143615
can be loaded now.

Author: huydhn

torchci/clickhouse_queries/recent_pr_workflows_query/query.sql

@@ -31,6 +31,8 @@ recent_prs AS (
     pull_request.base.'repo'.'full_name' = {repo: String}
     -- Filter pull request table to be smaller to amke query faster
     and pull_request.number in (select number from materialized_views.pr_by_sha where head_sha in (select head_sha from relevant_shas))
+    -- This query is used by Dr.CI, so we just want to update open PR(s)
+    and pull_request.state = 'open'
 )
 SELECT
   w.id AS workflowId,

torchci/lib/fetchPR.ts

@@ -47,11 +47,13 @@ export default async function fetchPR(
 
   // Ideally historicalCommits will be a superset of commits, but if there's a propagation delay with
   // getting the data to our database it may be missing recent commits for a bit.
-  if (shas.length == 0) {
+  if (shas.length === 0) {
     // If we got no data from our database, just use the commits from GitHub.
     shas = commits.map((commit) => {
       return { sha: commit.sha, title: commit.commit.message.split("\n")[0] };
     });
+  } else if (commits.length === 0) {
+    return { title, body, shas };
   } else {
     // For the very last sha, check to see if the shas themselves match as a proxy for detecting any missing commit.
     const lastCommit = commits[commits.length - 1];

torchci/pages/[repoOwner]/[repoName]/pull/[prNumber].tsx

@@ -120,8 +120,9 @@ function Page() {
 
   useEffect(() => {
     const selected = (sha ??
-      prData?.shas[prData.shas.length - 1].sha ??
-      "") as string;
+      (prData && prData.shas.length > 0
+        ? prData?.shas[prData.shas.length - 1].sha
+        : "")) as string;
     setSelectedSha(selected);
   }, [prData?.shas, sha]);
 
@@ -130,6 +131,7 @@ function Page() {
   if (prData === undefined) {
     return <div>Loading...</div>;
   }
+
   return (
     <div>
       <Stack
@@ -156,18 +158,23 @@ function Page() {
           repo={repoName as string}
         />
       </Stack>
-      <CommitHeader
-        repoOwner={repoOwner as string}
-        repoName={repoName as string}
-        prData={prData}
-        selectedSha={selectedSha}
-      />
-      <ErrorBoundary>
-        <CommitInfo
+      {selectedSha === "" && <div>Empty pull request without any commit</div>}
+      {selectedSha !== "" && (
+        <CommitHeader
           repoOwner={repoOwner as string}
           repoName={repoName as string}
-          sha={selectedSha}
+          prData={prData}
+          selectedSha={selectedSha}
         />
+      )}
+      <ErrorBoundary>
+        {selectedSha !== "" && (
+          <CommitInfo
+            repoOwner={repoOwner as string}
+            repoName={repoName as string}
+            sha={selectedSha}
+          />
+        )}
       </ErrorBoundary>
     </div>
   );

torchci/pages/api/drci/drci.ts

@@ -305,14 +305,17 @@ function removeFailureContext(failure: {
 async function getPRsWithPendingJobInComment(repo: String): Promise<number[]> {
   const query = `
 select
-    issue_url
+    issue_comment.issue_url
 from
     default.issue_comment final
+    join default.pull_request on issue_comment.issue_url = pull_request.issue_url
 where
     body like '<!-- drci-comment-start -->%'
     and match(body, '\\d Pending')
     and issue_comment.updated_at > now() - interval 1 month
-    and issue_url like {repo: String}`;
+    and issue_url like {repo: String }
+    and pull_request.state = 'open'
+`;
   const results = await queryClickhouse(query, { repo: `%${repo}%` });
   return results.map((v) => parseInt(v.issue_url.split("/").pop()));
 }