Good discussions so far. Frankly, I don't see the point of adding code to handle ENOMEM in some places but abort in other places. This may make sense only if we handle vast majority of failures but only a very few very rare cases where we abort.
I am inclined to believe that recovering from a vast majority of ENOMEMs is going to be complex. I am not sure how we are going to test all those cases as well. Even if we were to recover gracefully, what about other daemons that we depend on as someone pointed out? IMHO, any library should handle ENOMEMs and let the caller handle the failure, but as an application I am not sure if aborting is a bad idea. Anyway, I am all for abort() on ENOMEM with limiting our own memory usage by limiting the state/acl/inode cache entries etc. :-) Regards, Malahal. ------------------------------------------------------------------------------ _______________________________________________ Nfs-ganesha-devel mailing list Nfs-ganesha-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel
