Package ShotKit is in the index but not showing up

[finestructure]

There was an attempt to add the package to the index which actually succeeded - but only because the previous submission was a manual PR which left packages.json without a trailing newline. The new addition simply fixed that but didn't add the package (because it was already in the list):

The package has a valid Package.swift manifest, which can be dumped both on macOS and Linux. It also has tools-version 5.9. These are the two top reasons for a package being in the list but not showing up.

It must be failing in the analysis stage.

For reference: https://github.com/birangdev/ShotKit.git
and: https://swiftpackageindex.com/birangdev/ShotKit.git

[Arashk-A]

Package author here (submitted via PackageList#15117) — thanks for picking this up.

One thing that may be useful for the investigation: I tagged 1.0.1 on 2026-09-08, after this issue was filed, so there is a newer version for analysis to pick up if that helps re-trigger anything.

That release did not touch the manifest. Two example executable targets (CafeExportTool, ComparisonTool) imported AppKit unconditionally, which broke xcodebuild -scheme ShotKit-Package -destination 'generic/platform=iOS' build even though iOS is a declared platform; both are now behind #if os(macOS). I mention it only because I had initially assumed that was why the package was not appearing. I do not think it can be, since the compatibility builds run after analysis and a failing build should surface as a red matrix rather than a missing page — but flagging it in case the analysis stage does something platform-aware that I am not accounting for.

swift package dump-package succeeds on both tags. Happy to run anything locally or restructure the repo if that would help narrow it down.

[Arashk-A]

Gentle nudge on this one — is there anything I can provide from the package side to help narrow it down? Happy to cut a fresh tag, adjust the repo structure, or run any diagnostics locally if that would be useful.

[finestructure]

Hi @Arashk-A , I'm afraid there's nothing further you could be doing at this point. I'm about to take a closer look which part of our processing fails with your package and how. I'll post an update here asap!

[finestructure]

So the issue was that the package failed analysis. I can see this in the package's status which was analysis_failed, but we don't record more details there.

When running analysis locally, I could not reproduce any error, neither on macOS nor on the target Linux image (albeit the arm64 version).

Interestingly, when I then triggered a manual analysis run it worked fine:

[ INFO ] Analyzing (id: EA88481A-6284-48B4-9B76-34C5F8D1C38C) ... [component: analyze]
[ INFO ] Checkout directory: /checkouts [component: analyze]
[ INFO ] pulling https://github.com/birangdev/ShotKit.git in /checkouts/github.com-birangdev-shotkit [component: analyze]
[ INFO ] Updating 1 packages for stage 'analysis' [component: analyze]

We do continuously run analysis, so it's unclear to me why subsequent passes didn't clear the error. I'll need to check if there are any instances where an initial error or some other condition would prevent those runs.

Your package is available now but its data might grow stale if analysis fails again. I'll make note to check on it in the coming days but please also flag if you see anything that seems outdated!

[Arashk-A]

Thanks for digging into this, and for writing up what you found.

One data point that might help with the question of why subsequent passes
didn't clear the error: there was a new version in the middle of that window.
Timeline from my side:

• Sep 6: 1.0.0 tagged, package merged into the list
• Sep 8: 1.0.1 tagged, plus several commits to main
• Sep 15: your manual analysis run, after which it appeared

So there was both a new tag and new default-branch activity between the initial
failure and the manual trigger, and neither cleared it. If a new version would
normally re-drive analysis, that seems to point at the failed state being sticky
rather than at there being nothing new to pick up.

The page looks correct now and the badges resolve. I'll keep an eye on it and
flag here if the data goes stale.