On Thu, Dec 03, 2015 at 08:19:11PM +0100, Volker Lendecke wrote: > On Thu, Dec 03, 2015 at 06:08:15PM +0000, Jelmer Vernooij wrote: > > [+cc devel@lists.openchange.org] > > > > On Thu, Dec 03, 2015 at 01:43:52PM +0100, Volker Lendecke wrote: > > > Julien, a few months ago we spoke about the use of Samba's > > > DEBUG in OpenChange. On your website you state that > > > OpenChange is designed for exactly Samba 4.1.18, I'm > > > guessing that the DEBUG issue is one of the reasons for > > > that. 4.1.18 starts to smell, given that within the 4.1 > > > series it has been superseded by 3 releases and it's EOL > > > date is on the horizon. > > > > > > When reading doc/developer/logging.txt in your tree I get > > > the impression that the whole issue could be solved by > > > either extending our debug backends (see lib/util/debug.c) > > > with an OpenChange-specific mechanism. This would interely > > > be possible, we have a whole bunch of highly specific > > > backends like LTTNG and GPFS. OpenChange wouldn't make a > > > difference. > > > > > > If you don't want to extend debug.c, you might also use the > > > debug_set_callback() function and redirect the Samba RPC/NDR > > > debugs into your own mechanism in line. > > > > > > How can we go forward here? > > > > We've currently only finished stage (0), > > documented in doc/developer/logging.txt. > > > > This means that there are three steps still remaining: > > > > * having pidl avoid calls to DEBUG() but to something that is public > > > > * redirect DCE/RPC log output related to OpenChange to the > > openchange logging. We don't want to redirect *all* output from > > Samba (as OpenChange gets loaded into the daemon binary), just that > > related to the OpenChange DCE/RPC interfaces. > > Can't you solve that with debug classes? debug.c is pretty > flexible these days. As far as I can tell, those functions are not accessible to external users like OpenChange that build against Samba. They live in the same header and C file as DEBUG().
> > * having oc_log.c actually write to a OpenChange-specific log file, > > rather than simply calling DEBUG(). > That should be simple, right? Yes, but we should hold off on this until the other steps are done, lest we end up with logs related to OpenChange in two places. Cheers, Jelmer
signature.asc
Description: PGP signature
_______________________________________________ devel mailing list devel@lists.openchange.org http://mailman.openchange.org/listinfo/devel
