feat(crypto): Add EncryptionInfo to Decrypted to-device variant

[BillCarsonFr]

The ProcessedToDeviceEvent::Decrypted variant now also have an EncryptionInfo field.
Format changed from Decrypted(Raw<AnyToDeviceEvent>) to Decrypted { raw: Raw<AnyToDeviceEvent>, encryption_info: EncryptionInfo) }

It is needed for future work on widget in order to support encrypted to device properly viaadd_event_handler

This is a follow-up on

• crypto: Add variants for plain text and encrypted to-device events #4935

• refactor(crypto): Move session_id from EncryptionInfo to AlgorithmInfo as it is megolm specific #4981

• Public API changes documented in changelogs (optional)

Signed-off-by:

[codecov]

Codecov Report

Attention: Patch coverage is 93.93939% with 2 lines in your changes missing coverage. Please review.

Project coverage is 85.91%. Comparing base (0f877a3) to head (3bfc862).
Report is 38 commits behind head on main.

Files with missing lines	Patch %	Lines
...es/matrix-sdk-common/src/deserialized_responses.rs	80.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #5074      +/-   ##
==========================================
- Coverage   85.91%   85.91%   -0.01%     
==========================================
  Files         335      335              
  Lines       36468    36481      +13     
==========================================
+ Hits        31331    31342      +11     
- Misses       5137     5139       +2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[poljar]

Why the stringly typing? Curve25519Publickey is a Copy type so it's cheaper to copy it around, provides stronger typing and has a to_base64() method if you do need a string.

I feel like we've been here already once.

[BillCarsonFr]

It tried to keep it consistent with the other variant.
I suppose I have then to use serde(deserialize_with = "deserialize_curve_key", .. "serialize_curve_key"?
I'll do it too for MegolmV1AesSha2 to stay consistant

[BillCarsonFr]

Actually I am not sure about this change.
To do it, I need to add vodozemac as a dependency of matrix-sdk-common (it is not yet), would I need also to add some feature = "e2e-encryption" cfg?

[BillCarsonFr]

I had another look, and if I just refactor both Olm and Megolm variant to use Curve25519Publickey instead of strings, I'll need to

• make the common crate to depend on vodozemac
• Then for ser/deser compatibility (particularly for Megolm variant as part of TimelineEventKind) we need to move deserialize_curve_key and serialize_curve_key from crypto crate down to common crate (that will also bring back a dependency to zeroize).

This is touching several tests files. Looks like a candidate for a separate PR

[poljar]

What happens if someone puts a raw and encryption_info field into the JSON of the event?

It seems to me that the sender of the event, perhaps in collusion with the homeserver, might insert an encryption_info on their own.

Come to think of, why does this need to support Serialize and Deserialize? Because the widgets API will transfer this verbatim to the widget using JSON?

In any case, I think it's safer to not have a migration here since we don't seem to need it.

[BillCarsonFr]

Come to think of, why does this need to support Serialize and Deserialize?

I'll check but I think it mostly to export them to bindings?
But I agree that those should not be persisted, but we previously said that if we put something as serializable we should provide migration. I wish we had another trait for things that can be persisted

[BillCarsonFr]

So you are right.. Serialize and Deserialize are not used at all (I also checked in wasm bindings and it is not). Removing all that

[poljar]

Alright, I still have a couple of nits I would like to see fixed.

But I think we're good after that.

[poljar]

Looks good.