On Wed, Aug 27, 2014 at 03:18:46PM +0100, Bill Fischofer wrote: > I like that suggestion but perhaps ODP_ERR() is imply the name for > ODP_LOG_FATAL()? That way we'd have ODP_LOG() for normal/informational logs > and ODP_ERR() for really bad things. >
I'm OK with that. There is a slight complication that the ODP_ERR and ODP_DBG macros are currently used both internally and by the test/example applications, which I think is an error. Although given the suggestion to rename anything that doesn't form part of the normative API that would need to change anyway. Actually is anything that's currently in odp_debug.h part of the API? It doesn't look like it, other than as a dependency of other APIs, perhaps they should be tagged as @internal so they don't appear in the documentation. -- Stuart. > > On Wed, Aug 27, 2014 at 9:07 AM, Stuart Haslam > <[email protected]<mailto:[email protected]>> wrote: > On Wed, Aug 27, 2014 at 12:58:08PM +0100, Mike Holmes wrote: > > Summary of Platform Call discussion > > > > The exit call could be wrapped into ODP_ERR possibly so that at compile > > time other abortive action can be taken > > global_init could take a call back for absolutely fatal errors that ODP_ERR > > calls allowing the application to perform some action, it must not allow > > the app to continue running however > > > > I had assumed that ODP_ERR would be replaced once the new logging API > was agreed and merged. Why not combine this with the proposed new > ODP_LOG_FATAL() API?.. if there's no callback registered the default > implementation would print the message then call abort(). If there is a > callback registered it should not return (from _FATAL errors) but if it > does the logger implementation can prevent it returning to the > application. > > -- > Stuart. > _______________________________________________ lng-odp mailing list [email protected] http://lists.linaro.org/mailman/listinfo/lng-odp
