What should be done?

Right now we're using ./tmp/records that are created by the integration tests to increase the fuzzing corpus of sjson.

This is sub-optimal, because it seems like most of the documents are not "valid" data documents. The sjson fuzzing applies some document validation, so invalid documents are skipped.

After 35mins of integration testing, I got 3348 .bin entries in the ./tmp/records folder. 3 repetitive runs of fuzzing report 0 documents being added to the corpus. I double-checked that this is due to the fact that these documents can't pass the validation.

if doc.ValidateData() != nil {
	continue
}

We could try finding a better way to generate a corpus for sjson fuzzing purposes.

This is a separate issue from #1273, so it can be planned, analyzed, and estimated separately.

Where?

sjson package + something new that provides that new corpus.

Definition of Done

• all handlers updated;
• unit tests added/updated;
• integration/compatibility tests added/updated;
• spot refactorings done;
• user documentation updated or an issue to create documentation created;
• something else?