Client#sync fails with Net::ReadTimeout on large timeout values

[johannes-krude]

Client#sync takes a timeout parameter which is the amount of time to wait in case no new events are happening.

Net:HTTP#read_timeout is the amount of time until Net::ReadTimeout is raised in case the read syscall does not return. The default for this is 60 (seconds).

When calling Client#sync with a large timeout (e.g. 10 minutes), Net::ReadTimeout is raised before the server responds with an empty response.

As a workaround Client#sync can be called multiple times with a small timeout value.

[ananace]

The Matrix protocol itself works better when the client runs syncs more often, since that shrinks the window that the sync token might have to calculate. But if long-period syncs are necessary then you can modify that setting on the API object;

client = MatrixSdk::Client.new('https://matrix.example.com', read_timeout: 600)
# Or with an existing Client/API object;
client.api.read_timeout = 600

Bear in mind that even if your client is set to allow for a longer read timeout, the Matrix server on the other end might have a reverse proxy in front of it that could abort the connection as well.