fix(server): omit redundant tombstone reconciliation discards
This commit is contained in:
1 parent
9e7c409ba7
commit
86a9cfdd72
9 files changed
+228
-3
No files matched your search
@@ -56,6 +56,15 @@ servers send scripts in 16 KiB chunks, allowing pending uploads to reach older
|
||||
clients without lowering inbound validation limits. Large inline commands,
|
||||
stdin, and reconciliation responses still need the updated client reader.
|
||||
|
||||
A compact tombstone is already discarded command data. Servers with the
|
||||
tombstone-reconciliation correction validate that evidence but omit redundant
|
||||
discard instructions for it. Without that correction, a growing tombstone
|
||||
ledger alone can make the reconnect reply exceed an old client's 32 KiB reader
|
||||
and cause a persistent reconnect loop. Preserve the ledger; upgrade the client
|
||||
or apply the separately tested server correction. Genuine large action lists
|
||||
and other large messages still need a current client reader. This correction
|
||||
does not enable transport auto-chunking.
|
||||
|
||||
Development builds may negotiate `rvbox.agent.v1.chunked` in the WebSocket
|
||||
handshake. When both endpoints support it, messages automatically use bounded
|
||||
16 KiB packets and are reassembled before protocol handling. Old peers retain
|
||||
|
||||
Reference in new issue
Block a user