Good idea, some continuous integration system could make all of of our
lives easier. That said, it could make our lives harder, also.
Isn't there lots and lots of stuff like that already around?
Might Hudson [1] be the thing we want here?
It hooks into the RCS and builds the software for many previously
defined targets and then complaints to the culprits (and other
interested parties) via e-Mail if they break something. It is said to be
pretty extensible, too.
It's java software, and i've heard s.o. say it's an annoyingly slow
contraption, but then again, why should we care. Just have it running on
a separate box or some spare VM, as it's not time-critical at all.
Greetings,
Flo
[1] http://hudson-ci.org/
Daniel Golle wrote:
i'm day-dreaming about an automated testing environment for OpenWrt
for some days. automated testing with JUnit and ruby-unit has made my
life much better and the time-loss when developing on OpenWrt without
decent automated testing can be very frustrating.
sure, everybody can just setup a bunch of scripts to test-through
whatever, but sharing the test-cases and would be great.
therefore we need a common language to specify test-cases. i want to
share the ideas i had so far. i'd be glad to hear your opinion and
anything your feel being related to this idea.
why?
- increase quality
- have better ways to find, reproduce and communicate information about bugs
- keep us busy with developing, less time spending doing repetitive
things and waiting for results
how?
- have a test-host connected to one or more OpenWrt-based test-clients
(via serial and eventually using a reset-switch)
- let the host control and record what happens on the clients
- maybe each openwrt package can come with test-cases?
- maybe we can have a test-spec for each hardware target?
- possibility to specify multi-device-test cases, like have one be an
wifi ap and the other be wifi client
- have quantitative metrics and reports for some tests (like wifi
data-throughput)
- in case of a non-satisfying test-result, a bug-report can be
generate which, after approval, would be pushed into the bug-tracker
test-cases could be organized in an object-oriented fashion, i'm
thinking of an abstract testing-framework running on the test-host
which processes YAML or XML files holding the information of the tests
to be executed. i'd implement that in Ruby using ruby-unit.
dreams...
- have a hardware test-farm
- have a web-interface like tinderbox.mozilla.org
please share whatever comes up in you!
_______________________________________________
openwrt-devel mailing list
openwrt-devel@lists.openwrt.org
https://lists.openwrt.org/mailman/listinfo/openwrt-devel
_______________________________________________
openwrt-devel mailing list
openwrt-devel@lists.openwrt.org
https://lists.openwrt.org/mailman/listinfo/openwrt-devel
