SOLR-17876: Http2SolrClient: followRedirects=true flaky (#3505)

If someone creates an Http2SolrClient with followRedirects=true, and then uses this client to create future clients (via builder.withHttpClient), then redirect processing would be disabled on the shared instance (bad).

Author: dsmiley

solr/CHANGES.txt

@@ -263,6 +263,10 @@ Bug Fixes
 
 * SOLR-17871: Http2SolrClient wasn't honoring idle timeout configuration above 30 seconds -- a regression. (Thomas Wöckinger, David Smiley)
 
+* SOLR-17876: Http2SolrClient: followRedirects=true: if Http2SolrClient was created with followRedirects=true,
+  and then was used to create future clients (via builder.withHttpClient), then redirect processing was wrongly
+  disabled on the shared instance. (David Smiley)
+
 Dependency Upgrades
 ---------------------
 (No changes)

solr/solrj/src/java/org/apache/solr/client/solrj/impl/Http2SolrClient.java

@@ -152,11 +152,12 @@ protected Http2SolrClient(String serverBaseUrl, Builder builder) {
       this.httpClient = createHttpClient(builder);
       this.closeClient = true;
     }
+    // note: do not manipulate httpClient below this point; it could be a shared instance
+
     if (builder.listenerFactory != null) {
       this.listenerFactory.addAll(builder.listenerFactory);
     }
     updateDefaultMimeTypeForParser();
-    this.httpClient.setFollowRedirects(Boolean.TRUE.equals(builder.followRedirects));
     this.idleTimeoutMillis = builder.getIdleTimeoutMillis();
 
     try {
@@ -291,7 +292,7 @@ private HttpClient createHttpClient(Builder builder) {
     httpClient.setExecutor(this.executor);
     httpClient.setStrictEventOrdering(false);
     httpClient.setConnectBlocking(true);
-    httpClient.setFollowRedirects(false);
+    httpClient.setFollowRedirects(Boolean.TRUE.equals(builder.followRedirects));
     httpClient.setMaxRequestsQueuedPerDestination(
         asyncTracker.getMaxRequestsQueuedPerDestination());
     httpClient.setUserAgentField(new HttpField(HttpHeader.USER_AGENT, USER_AGENT));