Closed einhverfr closed 7 months ago
One small doubt I have which I have not been able to fully test is whether the LSN stuff properly moves forward with the guard destruction moved to another thread. For 1.0 I don't think this matters and we can find workarounds, but it would be a very good reason to come up with a fully async etcd driver.
I have now reviewed for missing dependency additions.
I am currently working on breaking the TestDecoding stuff off into a CPAN module. Maybe we want to just fix the current tests and then check this ne after it that is done (hopefully in the next couple days).
This now depends on onemoredata/bagger-overlay#4
This has required a number of changes including moving the replication callback to a new thread and fully rewriting the TestDecoding parser to use RecDescent.
Fixes #79
Some error handling improvements also, fixes #77