I was having weird issues with running Golang CI locally, and had to match the Go version.

Tests generated by gemini CLI

Assuming we still want to process all the blocks in a file, I think it'd be really good to have some test cases that include 2+ blocks

I wish I knew more about Coder templates. Do templates have asource field (or anything equivalent) that might need to be validated, too?

no its not applicable to templates

Do we want to swap the[^/]+ patterns for something even more specific? Right now, they allow any non-slash character, but don't we want to make sure the URLs don't have certain characters like underscores?

Assuming we still want to process all the blocks in a file, I think it'd be really good to have some test cases that include 2+ blocks

This is an edge case, but I feel like we want to handle cases where someone accidentally nests Terraform snippets inside each other like this:

```tf```tf

Edit: I think there's something we can salvage from this comment, but I realized in the second comment in the chain that it doesn't do everything we want

I don't know if this variable makes sense. With how it's set up right now, if we have at least one correct source in the README file, we'll automatically ignore all otherincorrect sources. And from a state modeling perspective, the variable also feels redundant, when a single slice should be able to give us all the info we need

What makes more sense to me is to:

1. HaveincorrectSources as the main validation state
2. Instead of having theactualSource == expectedSource check, only have aactualSource != expectedSource check. If that triggers, push the incorrect source to theincorrectSources slice. If things match, we'll do nothing, and let the code fall through to the rest of the loop
3. Once the main loop is done and we're done processing the lines, check if the slice is not empty. If it's not, return an error with a list of all sources that are incorrect

Actually, now that I'm thinking over the code more, I think I understand what the old code was trying to do, and why my suggestion doesn't actually work

I'm not super up to speed on our Terraform conventions, but if a module has a name likemodule "blah", would it be expected that the source URL should always have"blah" in it, too? If so, I think we'd need to check the name of each module block, and compare that against the source URL line

but if a module has a name like module "blah", would it be expected that the source URL should always have "blah" in it, too?

No it can be different. We just use the same name for convenience.

I think the test setup is pretty good. Just to account for the other comment I added, though, I feel like we should have one more test that validates what happens if you have a block with an incorrect body, and 2+ other blocks with correct bodies

Nit: I'm just now realizing how the sample MD contents are formatted, and I think I'd prefer for each one to be a single raw string. All the+ signs are a bit hard to read in the GitHub UI, and this is a case where all the spacing and lines do matter

I'm not sure why we wouldn't want to report all incorrect sources?

Nit: since this variable isn't used for a while, we can move it down, closer to where it's used