[log4net] exclusive lock on .NET Core 1.x and Linux

[Stefan Bodewig <bo...@apache.org>]

Hi all

it looks as if exclusive locking didn't work properly for
RollingfileAppender on Linux for our NET Core tests. This has also been
true for the MONO_2_0 builds ages ago. I haven't checked whether we are
hitting a known limitation so far.

I have disabled the test conditionallly on Linux and the Jenkins build
now finishes the tests on Linux.

Stefan

Re: [log4net] exclusive lock on .NET Core 1.x and Linux

[Dominik Psenner <dp...@gmail.com>]

On 28 Jan 2018 11:45 a.m., "Stefan Bodewig" <bo...@apache.org> wrote:

On 2018-01-21, Dominik Psenner wrote:

> Sometimes it is possible to configure the test runner so that it runs in
> process instead of spawning a new process. Would you like to investigate
on
> this?

I don't see any such option. We could re-enable more detailed logging
and try to figure out which of the tests crashes the containers in
Jenkins. So far I have been completely unable to reproduce the Jenkins
issue on Docker containers I've created locally.

Stefan


Same here. I'll escalate and forward a question to nunit-discuss asap.

Re: [log4net] exclusive lock on .NET Core 1.x and Linux

[Stefan Bodewig <bo...@apache.org>]

On 2018-01-21, Dominik Psenner wrote:

> Sometimes it is possible to configure the test runner so that it runs in
> process instead of spawning a new process. Would you like to investigate on
> this?

I don't see any such option. We could re-enable more detailed logging
and try to figure out which of the tests crashes the containers in
Jenkins. So far I have been completely unable to reproduce the Jenkins
issue on Docker containers I've created locally.

Stefan

Re: [log4net] exclusive lock on .NET Core 1.x and Linux

[Dominik Psenner <dp...@gmail.com>]

Sometimes it is possible to configure the test runner so that it runs in
process instead of spawning a new process. Would you like to investigate on
this? I had contact to Rob Prouse of the nunit developers. It might be a
good idea to crosspost to nunit-discuss.

On 21 Jan 2018 11:21 a.m., "Stefan Bodewig" <bo...@apache.org> wrote:

> On 2018-01-20, Stefan Bodewig wrote:
>
> > I have disabled the test conditionallly on Linux and the Jenkins build
> > now finishes the tests on Linux.
>
> But not reliably, there seems to be another test that crashes the test
> runner from time to time.
>
> https://developercommunity.visualstudio.com/content/
> problem/56318/the-active-test-run-was-aborted-reason-unable-to-c.html
>
> indicates there is no way for us to avoid the crashed test runner or to
> get decent information about which test is responsible for the crash (we
> may be able to see somthing if we increase verbosity again, oh my).
>
> There are reports of this problem being fixed in .NET Core 2.x but that
> won't help us.
>
> Unfortunately it doesn't seem to happen on the local docker image I've
> created for my own tests. This really is frustrating.
>
> Stefan
>

Re: [log4net] exclusive lock on .NET Core 1.x and Linux

[Stefan Bodewig <bo...@apache.org>]

On 2018-01-20, Stefan Bodewig wrote:

> I have disabled the test conditionallly on Linux and the Jenkins build
> now finishes the tests on Linux.

But not reliably, there seems to be another test that crashes the test
runner from time to time.

https://developercommunity.visualstudio.com/content/problem/56318/the-active-test-run-was-aborted-reason-unable-to-c.html

indicates there is no way for us to avoid the crashed test runner or to
get decent information about which test is responsible for the crash (we
may be able to see somthing if we increase verbosity again, oh my).

There are reports of this problem being fixed in .NET Core 2.x but that
won't help us.

Unfortunately it doesn't seem to happen on the local docker image I've
created for my own tests. This really is frustrating.

Stefan

Re: [log4net] exclusive lock on .NET Core 1.x and Linux

[Stefan Bodewig <bo...@apache.org>]

On 2018-01-20, Stefan Bodewig wrote:

> it looks as if exclusive locking didn't work properly for
> RollingfileAppender on Linux for our NET Core tests. This has also
> been true for the MONO_2_0 builds ages ago. I haven't checked whether
> we are hitting a known limitation so far.

https://github.com/dotnet/corefx/issues/22011

we use FileShare.Read which translates to no real locking at all on
Unix IIUC.

Stefan