Are the nonProxyHosts evaluated correctly in RestClientBuilderImpl?

[arne-reinhardt]

I'm having an issue with the way the http.nonProxyHosts Sytem Property is evaluated by org.jboss.resteasy.microprofile.clientRestClientBuilderImpl. I have thought of filing an Issue but I think this might be viewed as a question of design so let's discuss it.

Scenario:
I need to access Service my-pod.someserver.somecluster.mycompany.com
For some reason there are Proxy settings -Dhttp.proxyHost=proxy.mycompany.com -Dhttp.proxyPort=8080 -Dhttp.nonProxyHosts=*.mycompany.com|127.*|localhost

Problem:
Given the Implementation of RestClientBuilderImpl.getProxyHostsAsRegex() my Service call will be directed via the Proxy.
Why is this? Because said method converts the Pattern *.mycompany.com to a regex [A-Za-z0-9-]*\\.mycompany\\.com. This regex does not match my-pod.someserver.somecluster.mycompany.com. It will match no address containing a . that is not explicitely written. And of course it will not match 127.0.0.1. This behaviour also is different from what's stated on the Oracle Web Site.
The only solution is for me - or whoever is in control of the JVM arguments - to add *.someserver.somecluster.mycompany.com to the http.nonProxyHosts Property. Assuming there are a number of different testing environments this can amount to an awful lot of *.someotherserver.productioncluster.mycompany.com-stuff to configure - that's what wildcards were introduced for in the first place, I assume.

So wouldn't it make sense to change the implementation of the Pattern Matching around RestClientBuilderImpl.getProxyHostsAsRegex() and RestClientBuilderImpl.build() to be more useful?
A minimal change to make the implementation fit my needs might be to simply expand [A-Za-z0-9-]* to [A-Za-z0-9-\\.]*
But - frankly: the Pattern .*\\.mycompany\\.com would do the job perfectly well.

[jamezp]

What version RESTEasy and RESTEasy MicroProfile are you using? I thought this was fixed with #10, but we may have missed something. I do think RESTEasy should be duplicating what the JVM does. If not, IMO, it's a bug we should fix.

[arne-reinhardt]

We're using 1.0.0.Final of RESTeasy Microprofile and 5.0.1.Final of RESTeasy.

RESTeasy Microprofile 1.0.0.Final does contain Fix #10 so upgrading to a newer version would not help. Besides we're still bound to Java 8 currently - sad to say.

The issue in #10 as I understand it was that you could not directly use the nonProxyHosts parameter as a regex because it might be formally invalid. This has then been fixed to a regex that is formally valid but doesn't make sense in many cases.
Formal validity has been achieved by replacing . with \\. first and then replacing * with [A-Za-z0-9-]*
I think in step 2 you should rather be replacing * with .*
I see no advantage in the more complicated and less powerful expression you are currently using. Apart from maybe checking hostname restrictions that are none of your business at that point.

[jamezp]

This looks like an issue. I filed a issue for it. #143