feat(gax): add recoverable error query and buffer realignment loop

[whowes]

Introduces ChunkAttemptCallable with a query-and-realign loop when encountering recoverable protocol errors. Queries the server for the committed offset and adjusts the buffer window before resuming chunk transmission.

[gemini-code-assist]

Code Review

This pull request implements Category 2 (recoverable) error recovery for resumable uploads in GAX Java, updating ChunkAttemptCallable to query session status, realign the buffer window, and dispatch chunk uploads upon encountering recoverable errors or missing status headers. The review feedback highlights critical improvement opportunities in ChunkAttemptCallable, including addressing a potential infinite loop by correctly classifying query status failures under UploadCommand.QUERY rather than UPLOAD, and preventing race conditions by checking if the retrying future is cancelled immediately after setting inFlightFuture for both query and chunk upload futures.

[gemini-code-assist]

If the query status call fails with a non-transient error (such as 400 Bad Request or 403 Forbidden), classifying it under the chunk's upload command (e.g., UPLOAD) in UploadResultRetryAlgorithm would incorrectly classify it as RECOVERABLE (since 400 is recoverable for UPLOAD). This would trigger another recovery attempt, leading to an infinite loop of failing query status calls. We should classify the query failure using UploadCommand.QUERY and fail the attempt with a non-retryable exception if it is not transient.

          @Override
          public void onFailure(Throwable t) {
            UploadErrorCategory category = UploadErrorClassifier.classify(t, UploadCommand.QUERY);
            if (category == UploadErrorCategory.TRANSIENT) {
              failAttempt(attemptFuture, t);
            } else {
              failAttempt(
                  attemptFuture,
                  new IllegalStateException("Fatal error during query status: " + t.getMessage(), t));
            }
          }

[gemini-code-assist]

To prevent a race condition where the retrying future is cancelled concurrently before inFlightFuture is set, we should check if currentRetryingFuture is cancelled immediately after setting inFlightFuture and propagate the cancellation if so.

    this.inFlightFuture = queryFuture;
    if (currentRetryingFuture.isCancelled()) {
      queryFuture.cancel(true);
    }

[gemini-code-assist]

To prevent a race condition where the retrying future is cancelled concurrently before inFlightFuture is set, we should check if currentRetryingFuture is cancelled immediately after setting inFlightFuture and propagate the cancellation if so.

    this.inFlightFuture = chunkFuture;
    if (currentRetryingFuture.isCancelled()) {
      chunkFuture.cancel(true);
    }

[sonarqubecloud]

Quality Gate failed for 'gapic-generator-java-root'

Failed conditions
C Reliability Rating on New Code (required ≥ A)

See analysis details on SonarQube Cloud

Catch issues before they fail your Quality Gate with our IDE extension SonarQube for IDE

[sonarqubecloud]

Quality Gate failed for 'gapic-generator-java-root'

Failed conditions
0.0% Coverage on New Code (required ≥ 80%)
C Reliability Rating on New Code (required ≥ A)

See analysis details on SonarQube Cloud

Catch issues before they fail your Quality Gate with our IDE extension SonarQube for IDE