When configuring a custom event loop for the http client, it overrides a previously configured resolver, see reactor.netty.transport.ClientTransport#runOn
Expected Behavior
Keep the configured resolver when a custom event loop is used. Additionally, I would like to avoid instantiating a default resolver that's outside of the client control.
Actual Behavior
The resolver gets replaced by a default implementation.
Steps to Reproduce
@Test
public void repoCase() {
final EventLoopGroup eventLoopGroup = new NioEventLoopGroup(1);
HttpClient.create(connectionProvider)
.resolver(DefaultAddressResolverGroup.INSTANCE)
.runOn(eventLoopGroup); --> This creates a new resolver
}Possible Solution
reactor.netty.transport.ClientTransport#runOn
@Override
public T runOn(LoopResources loopResources, boolean preferNative) {
T dup = super.runOn(loopResources, preferNative);
CONF conf = dup.configuration();
if (conf.resolver != null) {
return dup;
}
if (conf.nameResolverProvider != null) {
conf.resolver = conf.nameResolverProvider.newNameResolverGroup(conf.loopResources(), conf.preferNative);
}
else {
conf.resolver = ClientTransportConfig.DEFAULT_NAME_RESOLVER_PROVIDER
.newNameResolverGroup(conf.loopResources(), conf.preferNative);
}
return dup;
}Your Environment
- Reactor version(s) used:
1.0.6
- Other relevant libraries versions (eg.
netty, ...):
- JVM version (
java -version):
- OS and version (eg.
uname -a):
When configuring a custom event loop for the http client, it overrides a previously configured resolver, see
reactor.netty.transport.ClientTransport#runOnExpected Behavior
Keep the configured resolver when a custom event loop is used. Additionally, I would like to avoid instantiating a default resolver that's outside of the client control.
Actual Behavior
The resolver gets replaced by a default implementation.
Steps to Reproduce
Possible Solution
reactor.netty.transport.ClientTransport#runOnYour Environment
1.0.6netty, ...):java -version):uname -a):