fix vcr

[schochastics]

Fix #177
Fix #174 eventually too
still some tests are failing locally

[schochastics]

@chainsawriot I know you cannot test this yourself but maybe you see what the problem is? Locally all tests pass. Must be something with the token?

[chainsawriot]

One explanation is the secret (or the access to secret). Let me check.

[chainsawriot]

@schochastics locally, you are using the dep version of vcr, correct?

ok, I got the answer from the last line of a cassette.

[chainsawriot]

I think I also need a(nother) Mastodon account to test this. The one instance that I still have an account (which I don't use) blocks all automatic agents. Thank you LLMs!

[schochastics]

ok, I got the answer from the last line of a cassette.

Ah true that could of course be an issue

[chainsawriot]

I think that's probably be at least one of the reasons, because I have the same error locally with the CRAN version.

[schochastics]

@sckott do you have any suggestions?

[chainsawriot]

@schochastics First, if I've installed the dev. version of vcr (remotes::install_github("ropensci/vcr@95ebedc65243ed4c8f7f4ce1d446730c4a52cbce")), I can get almost all tests passed. Except

rtoot/tests/testthat/test-auth_verify.R

Lines 4 to 9 in 379e4d0

	test_that("verify_envvar (Good case)", {
	## The cassette was created with a valid envvar
	vcr::use_cassette("envvar", {
	expect_error(capture_message(verify_envvar()), NA)
	})
	})

<vcr_unhandled/rlang_error/error/condition>
Error in `RequestHandlerHttr$new(req)$handle()`: Failed to find matching request in active cassette.
i Use `local_vcr_configure_log()` to get more details.
i Learn more in `vignette(vcr::debugging)`.

One interim solution is to modify our GHA to install dev version of vcr with remotes.

@sckott Do you have an ETA of when the new vcr will be on CRAN?

[chainsawriot]

Minor points about partial matching

rtoot/tests/testthat/test-get_account_auth.R

Line 118 in 379e4d0

x <- get_account_relationships(id = id, token = fake_token)

[sckott]

Confirming you're using dev vcr, correct?

wrt timeline, i'm working on getting vcr to cran now, BUT webmockr and crul need to get to cran first, and given how glacially slow CRAN submissions are, I bet it'll be 1-2 weeks before new vcr binaries are all up on CRAN

[sckott]

do you have any suggestions?

for something specific or just in general about the failing tests?

[chainsawriot]

rtoot/.github/workflows/R-CMD-check.yaml

Line 44 in b3980cd

extra-packages: any::rcmdcheck

We can try to add

github::ropensci/vcr (or even fix at a ref) to extra_packages as an experiment.

[sckott]

maybe #179

[schochastics]

do you have any suggestions?

for something specific or just in general about the failing tests?

Was not sure if we messed up something with vcr you might be able to catch.
I will merge #179 and see how it goes.It would be nice to have a stable GitHub version with vcr dev

[schochastics]

Ok just the envvar missing. I guess we could store that in a GIthub secret? But given that is only necessary for one test, we can skip it?

[chainsawriot]

@schochastics I am fine with skip CI.

[sckott]

@schochastics happy to help, but helpful if you can point me to a failing test that you think is vcr releated

Have you tried using local_vcr_configure_log() to diagnose issues (when that error message is shown)?

[schochastics]

@sckott I think we managed to fix it now, thanks! At the end I think it was not really related to vcr but data that was used in the tests not being available anymore on Mastodon

[sckott]

@schochastics great!