CC-753 upgrade tester utils

[libmartinito]

No description provided.

[linear]

CC-753 Upgrade tester_utils everywhere

To support extensions, we've had to make a breaking change to tester_utils. The PR is here: codecrafters-io/tester-utils#10.

We need to migrate over all course testers (and the "tester-template") to use this new structure.

Here's an example PR that does the migration for Redis: codecrafters-io/redis-tester#37. The description in that PR contains a rough list of steps that need to be taken:

• Run make update_tester_utils
• Resolve all the go errors, get the program to compile
• Run CODECRAFTERS_RECORD_FIXTURES=true make test
  • The only changes to fixtures should be the stage name (which the tester doesn't have access to anymore)
• Remove all usages of CODECRAFTERS_CURRENT_STAGE_SLUG
  • If you want a value for CODECRAFTERS_TEST_CASES_JSON that covers all stages, load a backup and run this command in core: puts Submission.where(course_stage: C("<course_slug>").last_stage).first.test_runs.first.test_cases_for_tester.to_json
• Upgrade to Go 1.20.x

The list of repositories that use tester-utils can be found here: https://github.com/codecrafters-io/tester-utils/network/dependents. interpreter-tester and dns-server-tester don't have any changes yet, so you can just copy over contents from tester-template once that's updated.

[libmartinito]

it seems that the CompareOutputWithFixture function in tester utils requires a NormalizeOutputFunc. is it better that this is always required? not sure how to normalize output for git, so i just returned the output

[rohitpaulk]

Yep, it's okay to just return the output if that works - some challenges will have randomness involved (like with TCP port numbers), that's where the normalization function is useful.

[libmartinito]

there aren't tests for some of the stages. should i make a linear test for that for each tester or no

[libmartinito]

appreciate if you could take a look at one instance before doing the other testers

[rohitpaulk]

there aren't tests for some of the stages. should i make a linear test for that for each tester or no

Nope, it's fine - we'd want to retain whatever tests we have, don't need to add new ones as part of this migration.

appreciate if you could take a look at one instance before doing the other testers

@libmartinito you've got failing tests - other than that looks good. Does make test_with_git work locally?

[rohitpaulk]

(comment added)

[libmartinito]

yep it works locally. the problem was the randomness in the file names, contents, hashed values. updated to be more deterministic.

Updated to get tests to pass

[rohitpaulk]

@libmartinito the randomness in tests is intentional. We do this to prevent users from accidentally passing the stage by hardcoding values. Let's find another way around it - check what other challenges do