Merge a697bb92c5 into 44dda08b13

* Avoid stack overflow when fetching deeply nested comments

* add test case

* reduce comment depth, add docs

* decrease

* reduce max comment depth to 50

* fmt

* clippy

* cleanup

Cargo.toml

@@ -90,7 +90,7 @@ lemmy_db_views = { version = "=0.19.6-beta.7", path = "./crates/db_views" }
 lemmy_db_views_actor = { version = "=0.19.6-beta.7", path = "./crates/db_views_actor" }
 lemmy_db_views_moderator = { version = "=0.19.6-beta.7", path = "./crates/db_views_moderator" }
 lemmy_federate = { version = "=0.19.6-beta.7", path = "./crates/federate" }
-activitypub_federation = { version = "0.6.0-alpha1", default-features = false, features = [
+activitypub_federation = { version = "0.6.0-alpha2", default-features = false, features = [
   "actix-web",
 ] }
 diesel = "2.1.6"

api_tests/src/comment.spec.ts

@@ -858,3 +858,26 @@ test("Dont send a comment reply to a blocked community", async () => {
   blockRes = await blockCommunity(beta, newCommunityId, false);
   expect(blockRes.blocked).toBe(false);
 });
+
+/// Fetching a deeply nested comment can lead to stack overflow as all parent comments are also
+/// fetched recursively. Ensure that it works properly.
+test("Fetch a deeply nested comment", async () => {
+  let lastComment;
+  for (let i = 0; i < 50; i++) {
+    let commentRes = await createComment(
+      alpha,
+      postOnAlphaRes.post_view.post.id,
+      lastComment?.comment_view.comment.id,
+    );
+    expect(commentRes.comment_view.comment).toBeDefined();
+    lastComment = commentRes;
+  }
+
+  let betaComment = await resolveComment(
+    beta,
+    lastComment!.comment_view.comment,
+  );
+
+  expect(betaComment!.comment!.comment).toBeDefined();
+  expect(betaComment?.comment?.post).toBeDefined();
+});

crates/api_crud/src/comment/create.rs

@@ -30,10 +30,9 @@ use lemmy_db_views::structs::{LocalUserView, PostView};
 use lemmy_utils::{
   error::{LemmyErrorExt, LemmyErrorType, LemmyResult},
   utils::{mention::scrape_text_for_mentions, validation::is_valid_body_field},
+  MAX_COMMENT_DEPTH_LIMIT,
 };
 
-const MAX_COMMENT_DEPTH_LIMIT: usize = 100;
-
 #[tracing::instrument(skip(context))]
 pub async fn create_comment(
   data: Json<CreateComment>,

crates/apub/src/api/user_settings_backup.rs

@@ -103,13 +103,16 @@ pub async fn import_settings(
   context: Data<LemmyContext>,
 ) -> LemmyResult<Json<SuccessResponse>> {
   let person_form = PersonUpdateForm {
-    display_name: Some(data.display_name.clone()),
-    bio: Some(data.bio.clone()),
-    matrix_user_id: Some(data.matrix_id.clone()),
+    display_name: data.display_name.clone().map(Some),
+    bio: data.bio.clone().map(Some),
+    matrix_user_id: data.bio.clone().map(Some),
     bot_account: data.bot_account,
     ..Default::default()
   };
-  Person::update(&mut context.pool(), local_user_view.person.id, &person_form).await?;
+  // ignore error in case form is empty
+  Person::update(&mut context.pool(), local_user_view.person.id, &person_form)
+    .await
+    .ok();
 
   let local_user_form = LocalUserUpdateForm {
     show_nsfw: data.settings.as_ref().map(|s| s.show_nsfw),
@@ -312,8 +315,9 @@ where
 #[expect(clippy::indexing_slicing)]
 mod tests {
 
-  use crate::api::user_settings_backup::{export_settings, import_settings, UserSettingsBackup};
+  use crate::api::user_settings_backup::{export_settings, import_settings};
   use activitypub_federation::config::Data;
+  use actix_web::web::Json;
   use lemmy_api_common::context::LemmyContext;
   use lemmy_db_schema::{
     source::{
@@ -401,45 +405,6 @@ mod tests {
     Ok(())
   }
 
-  #[tokio::test]
-  #[serial]
-  async fn test_settings_partial_import() -> LemmyResult<()> {
-    let context = LemmyContext::init_test_context().await;
-
-    let export_user =
-      create_user("hanna".to_string(), Some("my bio".to_string()), &context).await?;
-
-    let community_form = CommunityInsertForm::new(
-      export_user.person.instance_id,
-      "testcom".to_string(),
-      "testcom".to_string(),
-      "pubkey".to_string(),
-    );
-    let community = Community::create(&mut context.pool(), &community_form).await?;
-    let follower_form = CommunityFollowerForm {
-      community_id: community.id,
-      person_id: export_user.person.id,
-      pending: false,
-    };
-    CommunityFollower::follow(&mut context.pool(), &follower_form).await?;
-
-    let backup = export_settings(export_user.clone(), context.reset_request_count()).await?;
-
-    let import_user = create_user("charles".to_string(), None, &context).await?;
-
-    let backup2 = UserSettingsBackup {
-      followed_communities: backup.followed_communities.clone(),
-      ..Default::default()
-    };
-    import_settings(
-      actix_web::web::Json(backup2),
-      import_user.clone(),
-      context.reset_request_count(),
-    )
-    .await?;
-    Ok(())
-  }
-
   #[tokio::test]
   #[serial]
   async fn disallow_large_backup() -> LemmyResult<()> {
@@ -475,4 +440,33 @@ mod tests {
     LocalUser::delete(&mut context.pool(), import_user.local_user.id).await?;
     Ok(())
   }
+
+  #[tokio::test]
+  #[serial]
+  async fn import_partial_backup() -> LemmyResult<()> {
+    let context = LemmyContext::init_test_context().await;
+
+    let import_user =
+      create_user("hanna".to_string(), Some("my bio".to_string()), &context).await?;
+
+    let backup =
+      serde_json::from_str("{\"bot_account\": true, \"settings\": {\"theme\": \"my_theme\"}}")?;
+    import_settings(
+      Json(backup),
+      import_user.clone(),
+      context.reset_request_count(),
+    )
+    .await?;
+
+    let import_user_updated =
+      LocalUserView::read(&mut context.pool(), import_user.local_user.id).await?;
+    // mark as bot account
+    assert!(import_user_updated.person.bot_account);
+    // dont remove existing bio
+    assert_eq!(import_user.person.bio, import_user_updated.person.bio);
+    // local_user can be deserialized without id/person_id fields
+    assert_eq!("my_theme", import_user_updated.local_user.theme);
+
+    Ok(())
+  }
 }

crates/apub/src/protocol/objects/note.rs

@@ -20,10 +20,9 @@ use lemmy_db_schema::{
   source::{community::Community, post::Post},
   traits::Crud,
 };
-use lemmy_utils::error::LemmyResult;
+use lemmy_utils::{error::LemmyResult, LemmyErrorType, MAX_COMMENT_DEPTH_LIMIT};
 use serde::{Deserialize, Serialize};
 use serde_with::skip_serializing_none;
-use std::ops::Deref;
 use url::Url;
 
 #[skip_serializing_none]
@@ -58,9 +57,19 @@ impl Note {
     &self,
     context: &Data<LemmyContext>,
   ) -> LemmyResult<(ApubPost, Option<ApubComment>)> {
-    // Fetch parent comment chain in a box, otherwise it can cause a stack overflow.
-    let parent = Box::pin(self.in_reply_to.dereference(context).await?);
-    match parent.deref() {
+    // We use recursion here to fetch the entire comment chain up to the top-level parent. This is
+    // necessary because we need to know the post and parent comment in order to insert a new
+    // comment. However it can also lead to stack overflow when fetching many comments recursively.
+    // To avoid this we check the request count against max comment depth, which based on testing
+    // can be handled without risking stack overflow. This is not a perfect solution, because in
+    // some cases we have to fetch user profiles too, and reach the limit after only 25 comments
+    // or so.
+    // A cleaner solution would be converting the recursion into a loop, but that is tricky.
+    if context.request_count() > MAX_COMMENT_DEPTH_LIMIT as u32 {
+      Err(LemmyErrorType::MaxCommentDepthReached)?;
+    }
+    let parent = self.in_reply_to.dereference(context).await?;
+    match parent {
       PostOrComment::Post(p) => Ok((p.clone(), None)),
       PostOrComment::Comment(c) => {
         let post_id = c.post_id;

crates/db_schema/src/source/local_user.rs

@@ -14,11 +14,12 @@ use serde_with::skip_serializing_none;
 use ts_rs::TS;
 
 #[skip_serializing_none]
-#[derive(Clone, PartialEq, Eq, Debug, Serialize, Deserialize)]
+#[derive(Clone, PartialEq, Eq, Debug, Serialize, Deserialize, Default)]
 #[cfg_attr(feature = "full", derive(Queryable, Selectable, Identifiable, TS))]
 #[cfg_attr(feature = "full", diesel(table_name = local_user))]
 #[cfg_attr(feature = "full", diesel(check_for_backend(diesel::pg::Pg)))]
 #[cfg_attr(feature = "full", ts(export))]
+#[serde(default)]
 /// A local user.
 pub struct LocalUser {
   pub id: LocalUserId,

crates/utils/src/lib.rs

@@ -29,6 +29,8 @@ pub const CACHE_DURATION_FEDERATION: Duration = Duration::from_secs(60);
 
 pub const CACHE_DURATION_API: Duration = Duration::from_secs(1);
 
+pub const MAX_COMMENT_DEPTH_LIMIT: usize = 50;
+
 #[macro_export]
 macro_rules! location_info {
   () => {