fix: nil dereference in archive priority processing

[lczyk]

• Have you signed the CLA?

---

very niche, but the test by itself leads to a panic since we collect keys from yamlArchives but then index release.Archives with them

[lczyk]

sorry for the empty commit noise. the github workers were loosing their minds and i needed to retrigger a couple of times

[github-actions]

Command	Mean [s]	Min [s]	Max [s]	Relative
BASE	13.657 ± 0.084	13.553	13.783	1.00 ± 0.01
HEAD	13.624 ± 0.061	13.523	13.700	1.00

[upils]

Thanks @lczyk, good catch!

The test just need a bit of tuning and then it shall be good to move to core review.

[lczyk]

@upils both comments addressed. i wasn't 100% sure where to put the test, but i think i found a decent place.

also, i realized there is no equivalent test for v2/v3 so i've added that too (it's passing without any changes. the bug addressed here is with v1): df1581d

[upils]

Thanks for rework @lczyk. See my other comment regarding the tests.

[upils]

I believe this test is not tied to the "maintenance" feature. Unless I am missing something, it looks like the name (and the content of the chisel.yaml are making it difficult to actually know what is tested.

Also, to be inline with the last thing discussed in #267, I suppose a single test for V3 would be enough since the tested behavior is common for all formats.

[lczyk]

I believe this test is not tied to the "maintenance" feature.

ah, sorry, i thought Maintenance: ... meant tests of v1. changed in 9bbad89

it looks like the name (and the content of the chisel.yaml are making it difficult to actually know what is tested.

that's the smallest yaml which was causing the bug... 🤷‍♀️

i've added a comment to point the reader at the point of the test 768d763

I suppose a single test for V3 would be enough since the tested behavior is common for all formats.

the tests test different things. might be clearer now that the tests are next to one another + the name is reworded. also, importantly, without

- for archiveName := range yamlArchives {
+ for archiveName := range release.Archives {

only one of the tests fails ( only the v2-archive one ). hence, i believe there is a strong reason to keep both.

[upils]

@lczyk I think you missed this comment. If not, would you mind exposing why you think the second test must be kept?

[lczyk]

see above. sorry, stupid github decided to mark the comments as pending from some reason and never actually make them... 😠

[upils]

Thanks but see the previous comment I pinged you on (#269 (comment))

[upils]

Thanks for the improved test. I see now why 2 tests make sense.

See my other comment about tests location.