`oo7` fails to match secrets when searched attributes are a subset of stored attributes

[noxpardalis]

I was testing oo7 as a replacement for gnome-keyring on my system. I noticed that signing in to my GitHub account with zed (which uses the Secret Service API) would write a secret to the collection, but it couldn’t find that secret again after being closed and reopened.

When zed wrote the secret, the attributes looked like this:

{
  "url": "https://zed.dev",
  "username": "..."
}

However, it searched for the secret using only a subset of the attributes:

{
  "url": "https://zed.dev"
}

I'm not familiar enough with the FreeDesktop specification, but it seems reasonable to expect that items with a superset of the searched attributes should still match.

[A6GibKm]

Hello, you can use Bustle for example to see which calls to the secret services are being made and what values are being returned. Without knowing the API calls being made it is very hard to tell what went wrong.

Just knowing the called API methods and their arguments (without including sensitive data, if any) would help.

[noxpardalis]

Hello, you can use Bustle for example to see which calls to the secret services are being made and what values are being returned. Without knowing the API calls being made it is very hard to tell what went wrong.

Just knowing the called API methods and their arguments (without including sensitive data, if any) would help.

It's clear from the code here that the matches for attributes within a collection are for exact matches and not a subset:
https://github.com/bilelmoussaoui/oo7/blob/e00ed43016ce74f8b8914b326f91dbc39e6bd6ba/server/src/collection.rs#L232-L234

I opened PR #234 to change the behavior to match even if the searched attributes are a subset. Happy to supply more information if it's useful!

[bilelmoussaoui]

Nice! have you noticed any other issues? would appreciate future feedback about it :-)

[noxpardalis]

Nothing else atm (besides the stuff that still panics on the todos). I'll keep wiring it into my setup and will raise issues/open PRs if anything else pops up. Solid work so far, btw!

[bilelmoussaoui]

Per @A6GibKm tests, this seems to still be an issue. We will need actual tests to conform this works as expected, re-opening.