Dmitry Dygalo

Btw, in the sample in the original message I see `"s\ud8ac\udf0d>"`, which is 7 characters, not 4 as seen in the error message. Not sure how it could happen

It's indeed true; however, the strange part is that it's been serialized as `>` for JSON

I am going to approach it again the next weekend - there were a few hypotheses that I've tried but didn't have any success. > If it helps the backend...

hey! That's a lot of upvotes in such a short period of time! :) Right now, Schemathesis can output a VCR cassette with everything that was executed - `--report-vcr-path=cassette.yaml`, HAR...

There was a `replay` command that I've removed in Schemathesis v4, which somewhat reminds me of this scenario. However, to output just the cURL commands, you can write a CLI...

For stateful failures, I originally thought about something like this Showing them as a tree with complete sequences included felt more precise. Though it could be an overcomplication and just...

We have a minimal support for restricting lower-level values coming from the hypothesis-jsonschema level. Right now there is only some for string generation. In principle we wanted to rewrite it...

Thanks for reporting! That's indeed inconsistent - the coverage phase has a scenario where it generates test cases for methods that are not present in the schema for this endpoint;...

I think that's poorly documented and named - `unsupported_method` is a check that is independent of data generation. The generation-side counterpart is `phases.coverage.unexpected-methods` [here](https://schemathesis.readthedocs.io/en/stable/reference/configuration/#phasescoverageunexpected-methods). Hm, that's confusing too :(

@g3rda, thank you for opening this! I would also take a look at how other fuzzers do it. For example, `libfuzzer` can just run ` ` and it will run...