From 3498cb567287f1ecf74b67fdee2dc9307505c86e Mon Sep 17 00:00:00 2001
From: dawn <90008@klbr.net>
Date: Mon, 5 Oct 2026 21:11:29 +0300
Subject: [PATCH] [api] list a repo whose resync entry can't be read instead of
 failing the page

an unreadable resync entry, or one retrying at a time chrono can't hold, made
repo_row fail, and the listing collects into one Result, so every /repos page
holding that row 500'd and a cursor could never get past it. the row now lists
without resync and the log names the entry.
---
 src/control/repos/mod.rs | 57 ++++++++++++++++++++++++++++++++++++++--
 1 file changed, 55 insertions(+), 2 deletions(-)

diff --git a/src/control/repos/mod.rs b/src/control/repos/mod.rs
index f278456..acaead1 100644
--- a/src/control/repos/mod.rs
+++ b/src/control/repos/mod.rs
@@ -158,7 +158,6 @@ fn read_queues(state: &AppState, info: &mut RepoInfo, index_id: Option<u64>) ->
     else {
         return Ok(());
     };
-    let resync = rmp_serde::from_slice(&bytes).into_diagnostic()?;
     // a repo that never backfilled has no pds in its state yet, but resolving
     // it for the failed attempt left one in the identity cache
     let pds = info
@@ -166,7 +165,15 @@ fn read_queues(state: &AppState, info: &mut RepoInfo, index_id: Option<u64>) ->
         .clone()
         .or_else(|| state.resolver.cached_pds(&info.did));
     let host_failures = pds.map_or(0, |pds| state.throttler.snapshot(&pds).consecutive_failures);
-    info.resync = Some(ResyncInfo::new(resync, host_failures)?);
+    // one unreadable entry would fail every listing page it lands on, so the cursor
+    // could never get past it. the row lists without it and the log says why
+    match rmp_serde::from_slice(&bytes)
+        .into_diagnostic()
+        .and_then(|resync| ResyncInfo::new(resync, host_failures))
+    {
+        Ok(resync) => info.resync = Some(resync),
+        Err(e) => tracing::warn!(did = %info.did, err = %e, "skipping an unreadable resync entry"),
+    }
     Ok(())
 }
 
@@ -433,6 +440,52 @@ mod tests {
         assert!(ResyncInfo::new(state, 0).is_err());
     }
 
+    #[cfg(feature = "indexer")]
+    #[tokio::test]
+    async fn an_unreadable_resync_entry_does_not_stop_the_listing() -> miette::Result<()> {
+        use crate::types::{ResyncErrorKind, ResyncState};
+
+        let tmp = tempfile::tempdir().into_diagnostic()?;
+        let state = Arc::new(AppState::new(&crate::config::Config {
+            database_path: tmp.path().to_path_buf(),
+            ..Default::default()
+        })?);
+        let bad = Did::new_static("did:web:bad.example").into_diagnostic()?;
+        let garbled = Did::new_static("did:web:garbled.example").into_diagnostic()?;
+        let fine = Did::new_static("did:web:fine.example").into_diagnostic()?;
+        let mut batch = state.db.inner.batch();
+        for did in [&bad, &garbled, &fine] {
+            batch.insert(
+                &state.db.repos,
+                keys::repo_key(did),
+                db::ser_repo_state(&RepoState::backfilling())?,
+            );
+        }
+        let impossible = ResyncState::Error {
+            kind: ResyncErrorKind::Generic,
+            retry_count: 1,
+            next_retry: i64::MAX,
+        };
+        batch.insert(
+            &state.db.indexer.resync,
+            keys::repo_key(&bad),
+            rmp_serde::to_vec(&impossible).into_diagnostic()?,
+        );
+        batch.insert(
+            &state.db.indexer.resync,
+            keys::repo_key(&garbled),
+            b"not msgpack",
+        );
+        batch.commit().into_diagnostic()?;
+
+        let listed = ReposControl::new(state)
+            .iter(None)
+            .collect::<miette::Result<Vec<_>>>()?;
+        assert_eq!(listed.len(), 3);
+        assert!(listed.iter().all(|info| info.resync.is_none()));
+        Ok(())
+    }
+
     #[cfg(feature = "indexer")]
     #[tokio::test]
     async fn repo_info_shows_why_a_repo_waits_for_resync() -> miette::Result<()> {
-- 
2.54.0