System.NullReferenceException: Object reference not set to an instance of an object. at GitHubClient.GitHubApi.GetItem[T]

[danmoseley]

https://github.com/dotnet/aspire/actions/runs/13751888678/job/38453871652

network glitch? is there a missing error check/retry here?

[Pull Request #3656] Queued for prediction.
Unhandled exception. System.NullReferenceException: Object reference not set to an instance of an object.
   at GitHubClient.GitHubApi.GetItem[T](String githubToken, String org, String repo, UInt64 number, String itemQueryName) in /home/runner/work/aspire/aspire/src/GitHubClient/GitHubApi.cs:line 255
   at GitHubClient.GitHubApi.GetPullRequest(String githubToken, String org, String repo, UInt64 number) in /home/runner/work/aspire/aspire/src/GitHubClient/GitHubApi.cs:line 221
   at Program.<Main>$(String[] args) in /home/runner/work/aspire/aspire/src/Predictor/Predictor.cs:line 55
   at Program.<Main>(String[] args)
/home/runner/work/_temp/b62b21c0-ce8d-4efe-88f0-1a06fad6ebbe.sh: line 8:  1794 Aborted                 (core dumped) ./labeler-build/Predictor/Predictor --token "***" --repo "dotnet/aspire" --pull-model "labeler-cache/pull-model.zip" --pull-numbers "1-10000" --label-prefix "area-" --threshold 0.4 --default-label "needs-area-label"

[danmoseley]

same predicting iissues as pulls https://github.com/dotnet/aspire/actions/runs/13751885633/job/38453864396

Unhandled exception. System.NullReferenceException: Object reference not set to an instance of an object.
   at GitHubClient.GitHubApi.GetItem[T](String githubToken, String org, String repo, UInt64 number, String itemQueryName) in /home/runner/work/aspire/aspire/src/GitHubClient/GitHubApi.cs:line 255
   at GitHubClient.GitHubApi.GetIssue(String githubToken, String org, String repo, UInt64 number) in /home/runner/work/aspire/aspire/src/GitHubClient/GitHubApi.cs:line 218
   at Program.<Main>$(String[] args) in /home/runner/work/aspire/aspire/src/Predictor/Predictor.cs:line 23
   at Program.<Main>(String[] args)

[danmoseley]

https://github.com/dotnet/issue-labeler/blob/main/src/GitHubClient/GitHubApi.cs#L255

does not check GraphQLResponse.Errors before getting .Data ? If Errors is not empty perhaps log the error and retry a few times?

As for cause of the error, the above might help, but why would it fail in both jobs at the same time? Hmm, it was exactly 19:59:24 in both cases. Perhaps the server hiccuped.

[danmoseley]

It looks like this is one on the one other call to this API

https://github.com/dotnet/issue-labeler/blob/main/src/GitHubClient/GitHubApi.cs#L204

although, it doesn't retry.

[jeffhandley]

Thanks! Please re-run the job and see if it reoccurs. Either way, you're right that the error handling in this block needs improvement.

[danmoseley]

rerun
https://github.com/dotnet/aspire/actions/runs/13752130148/job/38454421527
https://github.com/dotnet/aspire/actions/runs/13752131341/job/38454423844

both failed immediately - I guess we need the error to know what happened.

Run ./labeler-build/Predictor/Predictor \
Loading issues model...
Issues prediction engine ready.
Unhandled exception. System.NullReferenceException: Object reference not set to an instance of an object.
   at GitHubClient.GitHubApi.GetItem[T](String githubToken, String org, String repo, UInt64 number, String itemQueryName) in /home/runner/work/aspire/aspire/src/GitHubClient/GitHubApi.cs:line 255
   at GitHubClient.GitHubApi.GetIssue(String githubToken, String org, String repo, UInt64 number) in /home/runner/work/aspire/aspire/src/GitHubClient/GitHubApi.cs:line 2[18](https://github.com/dotnet/aspire/actions/runs/13752130148/job/38454421527#step:4:19)
   at Program.<Main>$(String[] args) in /home/runner/work/aspire/aspire/src/Predictor/Predictor.cs:line 23
   at Program.<Main>(String[] args)

[danmoseley]

could config be corrupted somehow?, eg., repo is null.

[jeffhandley]

My guess is we exhausted the rate limit. Let's wait an hour and rerun the job. I'll use this issue to improve the error handling for this method similar to the method for downloading in bulk.

[jeffhandley]

#94 was submitted to add retry logic to the Predictor. #93 was also logged to capture room for improvement in the bulk prediction process.

The Onboarding page was updated to call out the API rate limit and give the recommendation of processing 2000 items per hour.

@danmoseley To validate that recommendation, I've processed all dotnet/aspire issues and pulls 2000 at a time successfully.