Am 15.03.2013 um 17:55 hat Laszlo Ersek geschrieben: > On 03/15/13 09:37, Kevin Wolf wrote: > > Am 14.03.2013 um 16:52 hat Laszlo Ersek geschrieben: > >> On 03/14/13 15:57, Kevin Wolf wrote: > >>> Signed-off-by: Kevin Wolf <kw...@redhat.com> > >>> --- > >>> After rebasing this I saw that Anthony already committed a fix that is > >>> very close to my v1. I don't intend to actually change that code, but as > >>> I've already done this, just for comparison what it would look like with > >>> error propagation. Is this what you meant? I find the result more > >>> confusing, to be honest. > >> > >> I think what I had in mind was: > >> - I was okay with the logic change you suggested in your v1, just > >> - turn *errp accesses into local_err accesses, > >> - when returning, propagate the latter to the former. > >> > >> The logic seemed OK, I just suggested to keep the massage internal to > >> the function, only try to propagate it outwards at return time. IOW, > >> never read *errp. > > > > So you would have used my local_err, but not ret_err? > > Something like that, yes. > > > I don't think that > > would make it much better, > > Not contesting that ;) > > > ret_err is actually the nice part. > > Anyway I'm not feeling strongly about this and I don't want to waste > your time with it. It was just a note in passing. (... Which I should > probably refrain from, lest I waste people's time.)
I'm not going to change this instance anyway now that Anthony pushed his own fix instead of mine. However this won't be the last time that I have to deal with an Error object, so I thought I'd check what is good practice. Seems no such thing has established yet, which is an answer, even though not the one I was hoping for. Kevin
