refactorings/preparations for introduce asymetrically encrypted keyupdates by hpk42 · Pull Request #8620 · chatmail/core

hpk42 · 2026-08-22T21:19:08Z

This PR does not change any behaviour for code on main but contains the refactoring parts for introducing asynchronously encrypted key updates in #8621

IT should not be merged without also merging the actual keyupdate branch.

hpk42 · 2026-08-22T21:54:27Z

+        // and nothing else should come out of it: no contact, no chat,
+        // and no `last_seen` update lighting up an online dot.
+        info!(context, "Report without message reference (TRASH).");
+        return trash().await;


not strictly neccessary but a short-cut to simplify reasoning about keyupdates in the future, to ensure the subsequent receive_imf code never sees this empty MDN.

This code is only for users who upgraded, so we still need to check that nothing really bad happens on versions that don't have this code (contacts/chats appearing etc., not just the green dot). Having this code before key updates may hide problems if something other than green dot on the address-contact appears as a result.

EDIT: it is tested with test_keyupdate_against_core_2_48_march_2026 in #8621

link2xt · 2026-08-26T17:33:14Z


+/// Returns the protected headers shared by encrypted messages
+/// that are not part of a chat.
+async fn non_chat_protected_headers(


All headers are protected in some way except for Chat-Is-Post-Message, so not sure what protected means here.

fixed the comment.

I guess the function should be renamed too, then? To non_chat_headers or headers_for_non_chat_msg?

I actually meant the function name, it is still _protected_. It's fine, i just don't know why it has _protected_ in the name.

link2xt · 2026-08-26T17:40:34Z

        Ok(mime_parser) => mime_parser,
    };

+    if !mime_parser.mdn_reports.is_empty()


I tried commenting this out (if false &&), the test test_mdn_without_message_reference fails then but only because a contact is created that is not visible.

Without this "address-contact" probably still gets a green dot, not a problem because the user needs to create a visible address-contact chat first to even see this contact.

No functional changes: Add a relay_addrs helper, share the protected headers and self-key rendering of non-chat messages, and move insert_into_smtp from securejoin to smtp.

…ages An unsigned message carries no intended recipient fingerprints, so recipients of an encrypted unsigned message learn nothing about other recipients from the PGP packets.

A report referencing no message can never be applied to one, so it must not create a contact, a chat or a `last_seen` update on its way to the trash.

Hocuri · 2026-08-26T07:52:46Z

 }

+/// Returns the protected headers shared by encrypted messages
+/// that are not part of a chat.


I didn't understand this comment until I looked at what the function is used for

Suggested change

/// that are not part of a chat.

/// Returns the protected headers shared by

/// encrypted messages that are not part of a chat,

/// i.e. key update messages and

/// symmetrically encrypted securejoin messages.

This is outdated, comment is already changed: #8620 (comment)

Hocuri · 2026-08-27T11:53:47Z

+/// Tests that a message encrypted without a signing key has no signature,
+/// and therefore no intended recipient fingerprints naming the other recipients.
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn test_encrypt_unsigned() {


This test is tautological, and doesn't test anything - it tests that if you don't pass any signing key to pk_encrypt, then it doesn't sign the message. And even if pk_encrypt somehow managed to sign the message with some dummy key (and include intended recipients), then the test would still pass, because pk_decrypt_and_validate ignores any signature that does not validate against sig_check_keyring.

I think the test can just be removed. I'm not sure if rpgp exposes a way to check if there is any signature inside a message; if so, then such a check could go into test_send_and_receive_keyupdate.

It is possible to check if the message is a signed message after decrypting or not, e.g. here:

core/src/mimeparser.rs

Line 462 in 322dd11

Some(pgp::composed::Message::Signed { reader, .. }) => reader.signature(0),

Decrypting the message should result in pgp::composed::Message::Literal message (because keyupdates are not compressed). But this is something for #8621 or PR on top of it.

test_send_and_receive_keyupdate already has this test, there is this code:

// A keyupdate is trashed on old cores because it's unsigned MDN // without a message reference. See also cross-core Python tests. let mail = mailparse::parse_mail(keyupdate.payload.as_bytes())?; let (mut decrypted, _fingerprint) = decrypt(bob, &mail).await?.unwrap(); // The next line is important: A key update message must NOT be signed, // as the signature might contain intended recipient fingerprints, // leaking all of the sender's contacts to all the other contacts. assert!(!decrypted.is_signed());

decrypted is a pgp::composed::Message

Hocuri · 2026-08-27T19:56:30Z

+/// Tests that an MDN referencing no message is trashed early:
+/// there is nothing it could ever be applied to,
+/// so it must not create a contact or a chat on the way.
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn test_mdn_without_message_reference() -> Result<()> {


Just from looking at the test and its documentation, it's unclear why we need this test, i.e. why it's important that an MDN referencing no message is trashed early.

I see two options:

adapt the test documentation to explain that this is important for key update messages

or replace it with a test testing that a key update message from an unknown contact doesn't create a contact or chat.

Just from looking at the test and its documentation, it's unclear why we need this test, i.e. why it's important that an MDN referencing no message is trashed early.

User-visible effect is that "last seen" is not updated for address-contact and address-contact does not get a green dot.
See also #8620 (comment)

Hocuri · 2026-08-27T20:02:39Z


+/// Returns the protected headers shared by encrypted messages
+/// that are not part of a chat.
+async fn non_chat_protected_headers(


I guess the function should be renamed too, then? To non_chat_headers or headers_for_non_chat_msg?

hpk42 mentioned this pull request Aug 22, 2026

introduce asymetrically encrypted keyupdate messages informing contacts about relay changes #8621

Open

hpk42 changed the title ~~refactorings/preparations for introduce asynchronously encrypted keyupdates~~ refactorings/preparations for introduce asymetrically encrypted keyupdates Aug 22, 2026

hpk42 force-pushed the hpk/prepare-keyupdates branch from a871291 to 7c7c14e Compare August 22, 2026 21:44

hpk42 commented Aug 22, 2026

View reviewed changes

hpk42 force-pushed the hpk/prepare-keyupdates branch from 7c7c14e to 715e1cc Compare August 22, 2026 22:02

j-g00da reviewed Aug 24, 2026

View reviewed changes

Comment thread src/pgp.rs Outdated

Hocuri requested review from Hocuri and link2xt August 26, 2026 07:55

link2xt approved these changes Aug 26, 2026

View reviewed changes

hpk42 added 3 commits August 27, 2026 14:23

refactor: extract shared pieces for non-chat messages

df5e3b0

No functional changes: Add a relay_addrs helper, share the protected headers and self-key rendering of non-chat messages, and move insert_into_smtp from securejoin to smtp.

feat: allow to not sign asymmetrically encrypted multi-recipient mess…

38fb022

…ages An unsigned message carries no intended recipient fingerprints, so recipients of an encrypted unsigned message learn nothing about other recipients from the PGP packets.

fix: trash early MDNs that reference no message

36c0231

A report referencing no message can never be applied to one, so it must not create a contact, a chat or a `last_seen` update on its way to the trash.

hpk42 force-pushed the hpk/prepare-keyupdates branch from 715e1cc to 36c0231 Compare August 27, 2026 12:54

link2xt reviewed Aug 27, 2026

View reviewed changes

Comment thread src/pgp.rs

link2xt reviewed Aug 27, 2026

View reviewed changes

Comment thread src/pgp.rs

Hocuri approved these changes Aug 28, 2026

View reviewed changes

link2xt mentioned this pull request Aug 28, 2026

refactor: import Autocrypt-Gossip keys without checking the addresses #8638

Open

-/// that are not part of a chat.
+/// Returns the protected headers shared by
+/// encrypted messages that are not part of a chat,
+/// i.e. key update messages and
+/// symmetrically encrypted securejoin messages.

Uh oh!

Uh oh!

Conversation

hpk42 commented Aug 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

link2xt Aug 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

link2xt Aug 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

hpk42 commented Aug 22, 2026 •

edited

Loading

link2xt Aug 26, 2026 •

edited

Loading

link2xt Aug 28, 2026 •

edited

Loading