question about logger tests

question about logger tests

[Ruediger Meier]

Hi,

Our logger tests fail if it can't connect to /dev/log. I'd like to fix 
that case for the test-suite but have some questions:

1. It was confusing for me to figure out the actual
   problem. Maybe --no-act and/or --stderr should imply
   that --socket-errors=auto turns error printing on?

2. Alternatively we could use --socket-errors=on for all tests.

3. Couldn't we fix --no-act to not need an open /dev/log at all?

cu,
Rudi

Re: question about logger tests

[Karel Zak]

On Thu, Mar 26, 2015 at 02:50:33PM +0100, Ruediger Meier wrote:
> Our logger tests fail if it can't connect to /dev/log. I'd like to fix 
> that case for the test-suite but have some questions:
> 
> 1. It was confusing for me to figure out the actual
>    problem. Maybe --no-act and/or --stderr should imply
>    that --socket-errors=auto turns error printing on?

Hmm.. probably good idea.

> 2. Alternatively we could use --socket-errors=on for all tests.
> 
> 3. Couldn't we fix --no-act to not need an open /dev/log at all?

But then it will introduce another fragility, complexity and
difference between test (--no-act) and non-test mode. I see for
example "if(ctl->fd < 0)" in code. Now it really skips write() only.

    Karel

-- 
 Karel Zak  <kzak@redhat.com>
 http://karelzak.blogspot.com

Re: question about logger tests

[Ruediger Meier]

On Friday 27 March 2015, Karel Zak wrote:
> On Thu, Mar 26, 2015 at 02:50:33PM +0100, Ruediger Meier wrote:
> > Our logger tests fail if it can't connect to /dev/log. I'd like to
> > fix that case for the test-suite but have some questions:
> >
> > 1. It was confusing for me to figure out the actual
> >    problem. Maybe --no-act and/or --stderr should imply
> >    that --socket-errors=auto turns error printing on?
>
> Hmm.. probably good idea.

I'll send a patch.

> > 2. Alternatively we could use --socket-errors=on for all tests.
> >
> > 3. Couldn't we fix --no-act to not need an open /dev/log at all?
>
> But then it will introduce another fragility, complexity and
> difference between test (--no-act) and non-test mode. I see for
> example "if(ctl->fd < 0)" in code. Now it really skips write() only.

You are right. If there would be a real-life use case for --no-act then 
not connecting /dev/log could be an optimization but I guess there is 
no such use case except our tests.

So how could we skip logger tests safely if /dev/log does not work?

cu,
Rudi

Re: question about logger tests

[Karel Zak]

On Fri, Apr 03, 2015 at 12:10:57AM +0100, Ruediger Meier wrote:
> So how could we skip logger tests safely if /dev/log does not work?

Would be enough to test by

    socat - UNIX-CONNECT:/dev/log

connect to the socket?

    Karel

-- 
 Karel Zak  <kzak@redhat.com>
 http://karelzak.blogspot.com

Re: question about logger tests

[Ruediger Meier]

On Tuesday 07 April 2015, Karel Zak wrote:
> On Fri, Apr 03, 2015 at 12:10:57AM +0100, Ruediger Meier wrote:
> > So how could we skip logger tests safely if /dev/log does not work?
>
> Would be enough to test by
>
>     socat - UNIX-CONNECT:/dev/log
>
> connect to the socket?

Well, using socat we can also use another socat listening socket instead 
of /dev/log at all. BTW we wouldn't even need --stderr and --no-act to 
check the written data.

I have done this in my github clone, branch "tests-logger"
https://github.com/rudimeier/util-linux/commits/tests-logger
(Not ready for merge yet.)

cu,
Rudi