[Dnsmasq-discuss] dnsmasq continuous integration

Geert Stappers stappers at stappers.nl
Fri Jan 5 07:53:10 GMT 2018


On Thu, Dec 14, 2017 at 09:27:34PM +0000, Simon Kelley wrote:
> On 13/12/17 10:10, Petr Men?ík wrote:
> > Hello everyone.
> > 
> > I maintain dnsmasq in Fedora and Red Hat Enterprise Linux.
> > 
> > We build for different sets of architectures and have some tests for
> > several packages. Dnsmasq is used for libvirt and network manager as a
> > dependency. In short, dnsmasq is important to us.
> > 
> > We lack something that we could attach after every build to test dnsmasq
> > is not (completely) broken. I do not know of any tests included in
> > dnsmasq repository. Is there any external repository that can be used to
> > validate dnsmasq still behaves properly? Is anyone using some kind of
> > continuos integration to ensure new build work at least the same as the
> > former one? I myself already made patches that broke some architectures
> > and some not. Such things are not easy to discover.
> > 
> > I currently assume no one has open source tests that can be used to
> > verify dnsmasq behavior right now. I would be glad if that was not true.
> > In the other case I would like to start some basic test suite, that can
> > be run to validate new build. I would like to make something useful that
> > could be merged into the repository sometime in the future.
> > 
> > I have a couple of questions:
> > * Do you know good and powerful enough framework to write such tests?
> > * Is there interest to cooperate on test suite? I think we could all
> > benefit from this, making dnsmasq more reliable.
> > * What language and library should be used for tests writing?
> > * Which tools would be useful for testing?
> > 
> > My kind of requirements:
> >   - setup support of temporary addresses, network namespaces or
> > containers would be useful
> >   - I would like to avoid reinventing the wheel, starting with bash
> > scripts, that would be simple at the start and hell to maintain later
> >   - I think scripting languages are more suited for complicated test
> > setup with more than one daemon instance
> >   - Preferred language would be python for me. Avocado [1] was
> > recommended to me. It is packaged in Fedora, but not in Debian. Because
> > Simon is Debian packager, I think something with good support on Debian
> > should be chosen. Do you know something?
> >   - check library [2] seems interesting. I am afraid current code would
> > be not easily broken into unit tests written in C
> >   - tests can be started as a single test or set of tests, failure
> > should be reported for each single test separately
> >   - dig would be useful for dns queries. ldns-testns can be quite useful
> > for special upstream DNS servers.	
> > 
> > What do you think? Any opinions would be appreciated.
> > 
> 
> I don't know of any such testing system, and I don't use one at present,
> the best I can do is fairly strict dogfood testing: the latest code is
> always running in the network router chez Kelleys.
> 
> If such a thing can be produced, I would be very interested in running it.


I read that as   "patches welcome"    :-)

 
 [1] https://github.com/avocado-framework/avocado
 [2] https://libcheck.github.io/check/


Groeten
Geert Stappers
-- 
Leven en laten leven



More information about the Dnsmasq-discuss mailing list