On Wed, 2008-04-30 at 14:14 +0100, Christine Caulfield wrote: > Steven Dake wrote: > > On Mon, 2008-04-28 at 11:05 +0100, Christine Caulfield wrote: > >> This is a patch that adds some standalone code to libconfdb, so it can > >> examine and ultimately change, the configuration without aisexec running. > >> > >> A small number of things I'm unsure about.. > >> > >> - it automatically falls back to standalone if it can't connect to > >> aisexec. That might not be the right thing to do, opinions welcome. > > > > > > Revised patch attached > > > > standalone mode should only be used when the environment variable for > > the configuration engine is set. If it isn't set, then it should return > > an error standard for aisexec not running. > > > >> - any program accessing confdb needs to be linked -rdynamic so that > >> several objdb/config symbols can be resolved. There might be ways around > >> this though I'm not sure how nice they are! > > > > Yikes > > Looking at it, I can't see any reasonable way round this. The lcrso > system requires that the registration code for the lcrso is accessible > from the loading program. > > Without -rdynamic, a program using libconfdb will give the following > error when stanadalone access is requested: > > /usr/libexec/lcrso/aisparser.lcrso: open failed: > /usr/libexec/lcrso/aisparser.lcrso: undefined symbol: lcr_component_register > > Normal access via aisexec doesn't need it though. > ok this is fine.
> > >> - Some config options I haven't got to work yet. In particular the new > >> cman preconfig module calls logsys, and I'm undecided about the best way > >> to tackle this. I think maybe the callers ought also to link with > >> logsys, but a quite test seems to make this difficult for some reason. > >> > >> > > > > The configuration engine plugins shouldn't be dependent upon logsys or > > any other third party library. > > OK, so the cman ones need logsys ripping ouy and replacing with syslog > ... there's progress ;-) > > The CMAN configuration engine shouldn't use logsys or any type of printf logging at all. Instead it should return its error in the proper api. The key is to separate the binaries so there is a cman_cfeng and a cman lcrso files. Doesn't this make sense? > Chrissie > _______________________________________________ > Openais mailing list > Openais@lists.linux-foundation.org > https://lists.linux-foundation.org/mailman/listinfo/openais _______________________________________________ Openais mailing list Openais@lists.linux-foundation.org https://lists.linux-foundation.org/mailman/listinfo/openais
