fix: Fix a possible NULL PTR after introduced timeout on waitForDone #1638
Conversation
…r we back out with a timeout on waitForDone
product-auto-label
bot
added
size: s
| @@ -573,6 +573,9 @@ private void requestCallback(AppendRowsResponse response) { | |||
| conectionRetryCountWithoutCallback = 0; | |||
| } | |||
| requestWrapper = pollInflightRequestQueue(); | |||
| if (requestWrapper == null) { | |||
There was a problem hiding this comment.
If we haven't cleaned up and this happens, we should throw an exception
Maybe instead of using null, check if we've cleaned up, and don't call the method (since this is under lock anyways)?
There was a problem hiding this comment.
If we have cleaned up, then all inflight requests are all set with an exception already.
I don't think we should throw on requestCallback?
There was a problem hiding this comment.
I meant the opposite: if we haven't cleaned up.
If during normal operations we get a callback but there's nothing in the queue, that should never happen, so we should throw (or somehow error out)
There was a problem hiding this comment.
Updated the code, hope it is more straightforward.
There was a problem hiding this comment.
Yeah it is, but I was thinking we'd also have
doneCallback {
this.done = true;
}
and for your change in requestCallback:
if (!this.done && this.inflightRequestQueue.isEmpty()) {
// Handle this error case!
}
But since it's in the callback, that's probably more complicated than I originally thought... Should we handle that in a followup?
There was a problem hiding this comment.
There is no super good place for done, since the lock is done inside of the cleanupInflightQueue.
There was a problem hiding this comment.
Added an error message, without a response, user will probably not be able to get the actual exception.
|
@yirutang |
yirutang
added
the
owlbot:run
gcf-owl-bot
bot
removed
the
owlbot:run
🤖 I have created a release *beep* *boop* --- ## [2.13.0](v2.12.2...v2.13.0) (2022-05-05) ### Features * add support to a few more specific StorageErrors for the Write API ([#1563](#1563)) ([c26091e](c26091e)) * next release from main branch is 2.12.2 ([#1624](#1624)) ([b2aa2a4](b2aa2a4)) ### Bug Fixes * A stuck when the client fail to get DoneCallback ([#1637](#1637)) ([3baa84e](3baa84e)) * Fix a possible NULL PTR after introduced timeout on waitForDone ([#1638](#1638)) ([e1c6ded](e1c6ded)) ### Dependencies * update dependency com.google.cloud:google-cloud-bigquery to v2.10.10 ([#1623](#1623)) ([54b74b8](54b74b8)) * update dependency org.apache.avro:avro to v1.11.0 ([#1632](#1632)) ([b47eea0](b47eea0)) ### Documentation * **samples:** update WriteComittedStream sample code to match best practices ([#1628](#1628)) ([5d4c7e1](5d4c7e1)) * **sample:** update WriteToDefaultStream sample to match best practices ([#1631](#1631)) ([73ddd7b](73ddd7b)) --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please).
This PR:
#1637
introduced a timeout for waitForDone. So we potentially could have drained the inflight requests when a RequestCallback comes back. Add process for null when peak the inflight queue.