Make sure very large changesets in bootstraps do not crash the sync client #6220
Conversation
src/realm/sync/client.cpp
Outdated
| } | ||
| catch (const LogicError& ex) { | ||
| if (ex.kind() == LogicError::binary_too_big) { | ||
| IntegrationException ex(ClientError::bad_changeset, |
There was a problem hiding this comment.
You can even throw bad_changeset_size if you wanted it since there is both a ClientError and ProtocolError and we can translate it before sending to the server.
|
@ironage , could you take a look while @danieltabacaru is out? |
There was a problem hiding this comment.
This seems fine as a stopgap fix, but it is still bad news that the server can sometimes cause this unrecoverable error on a client. AFAIK, core doesn't actually need to enforce the size limit in this case because it is the history. IIRC Realm can store arbitrary length binaries and the only reason we restrict the length is because we want to respect MongoDB's server side limit. I'd expect that if we used a different API to set the binary which didn't do the size check, everything should just work?
|
Yeah, there's a server-side fix for this captured in BAAS-19105 and this is a stop-gap fix. |
What, How & Why?
This fixes the sync client so that if it receives a FLX bootstrap changeset larger than what can fit into a BinaryColumn, it treats that as a bad_changeset_size error and fails gracefully rather than crashing the process.
☑️ ToDos