Hello,
I don't mind having separate profile for milo based tests, we actually have some conditional activators in these (i.e. system kind for OpcuaPlcDriverTest), switching to profile will simplify control over these. If ever these tests become stable enough we can re-consider switching profile to be active by default.

Cheers,
Łukasz

On 25.09.2023 13:00, Lukas Ott wrote:
+1 for disabling Milo, we should be able to build with docker compose up
without running into these issues.

If someone needs Milo test then actively enabling them seems like a
reasonable plan.

Lukad

Christofer Dutz <christofer.d...@c-ware.de> schrieb am Mo., 25. Sept. 2023,
10:59:

Hi Lukasz,

Well … we need a test-suite that runs reliably … in the past I usually ran
into issues with Milo not running on various configurations.
Admittedly I’m just no longer willing to invest the little time into
fixing issues with this bit of software.

For example, running “docker compose up” currently fails because in docker
something seems to fail to startup, which doesn’t fail when running on the
hardware directly.

I’m happy for a profile “option-with-milo-tests” that actively enables
Milo-based tests.
(I would insist on actively enabling it instead of actively disabling it
as otherwise people will continuously report problems with it)

Chris



Von: Łukasz Dywicki <l...@code-house.org>
Datum: Montag, 25. September 2023 um 10:49
An: dev@plc4x.apache.org <dev@plc4x.apache.org>
Cc: Patryk Gała <patryk.gala....@gmail.com>
Betreff: Re: Convert the Milo-based test suite into one using our
integration-test framework?
Milo is a bit painful, but it does more than we currently can do in our
opcua implementation. Also, from basic look I had into why some of tests
fail, it looks primarily like resource lookup issue. For example I
spotted a swallowed null pointer exception which comes from our code,
not Milo.

We will have difficult times making test framework for testing security
with certificates and Java crypto apis. Please hold on with major
changes in this area, as we intend using it to automate testing of UA
security policies (sign, sing&encrypt) using various modes (basic rsa,
basic aes etc.).
Back then Patryk had an idea of making a "test container" for Milo which
would separate it into its own process. Obviously it will need a bit of
work and will become a prerequisite to eventual test refactoring,
however it will make it (fail) same way for everyone. ;)

Best,
Łukasz

On 25.09.2023 10:38, Christofer Dutz wrote:
Hi all,

I know that I have spent many days tracking down oddness of problem with
OPC-UA tests, that all tracked down to Milo behaving oddly on various
platforms (Doesn#t start in Parallells VMs, has issues being run in docker,
…).
When building PLC4X I put a lot of effort into our integration-test
framework, which should allow testing our drivers on various platforms and
languages without needing a real (or simulated) device.

I think we should translate the Milo based tests into ones using our
DriverTestsuite framework, as admittedly I’m fedup with Milo …


Chris




Reply via email to