[pulseaudio-discuss] [PATCH] pulsecore: Introduce PULSE_LOG_JOURNAL environment variable

[Tanu Kaskinen]

On Wed, 2015-09-16 at 12:28 +0200, Ahmed S. Darwish wrote:
> [ Tanu, your MUA sometimes drops CCs. I'm the target this time ;-) ]

It's Mailman that randomly drops them (it really seems random - I
haven't figured out the logic). I got two mails, one directly from
David to my inbox, and one via Mailman to my pulseaudio-discuss folder.
The one from Mailman didn't have you in CC. I try to reply to the
direct mails for this reason, but occasionally I forget and reply to
the Mailman version.

> On Wed, Sep 16, 2015 at 12:51:29PM +0300, Tanu Kaskinen wrote:
> > On Wed, 2015-09-16 at 11:36 +0200, David Henningsson wrote:
> > > Due to this patch, my build (of -next) is now broken:
> > > 
> > > ../../src/pulsecore/log.c: In function 'init_defaults':
> > > ../../src/pulsecore/log.c:298:31: error: 'PA_LOG_JOURNAL'
> > > undeclared 
> > > (first use in this function)
> > >               target_override = PA_LOG_JOURNAL;
> > >                                 ^
> > > 
> > > ...probably because I don't HAVE_SYSTEMD_JOURNAL.
> > > 
> > > I kind of feel it would be better to have PA_LOG_JOURNAL always
> > > defined 
> > > than to add ifdefs every where we use it, but I'm open for either
> > > option. Any opinions?
> > 
> > So this failing code is in init_defaults():
> > 
> >         if (getenv(ENV_LOG_JOURNAL)) {
> >             target_override = PA_LOG_JOURNAL;
> >             target_override_set = true;
> >         }
> > 
> > In this case I feel it's better to just #ifdef that code out. Even
> > if
> > PA_LOG_JOURNAL was defined, I think this code should be compiled
> > out,
> > because I don't think we want to set target_override when the
> > target in
> > question isn't supported.
> > 
> 
> Shouldn't we also add a small warning to the user in this case?
> Something like:
> 
>         pa_log_warn("Cannot send logging messages to the jornal; "
>                     "systemd journal support was not compiled");

Yes, that would make sense. The message should also mention that the
request to use the journal came via an environment variable.

-- 
Tanu