On Mon, Mar 25, 2019 at 10:36:26AM +0100, Enrico Weigelt, metux IT consult wrote: > On 22.03.19 20:06, Andy Shevchenko wrote: > > On Fri, Mar 22, 2019 at 07:32:28PM +0100, Enrico Weigelt, metux IT consult wrote: > >> On 21.03.19 10:23, Andy Shevchenko wrote: > >> > >>> ...and on top of that GPIO sysfs interface is deprecated. > >> > >> I don't like the idea of deprecating this. It might not be enough for > >> all usecases, but for a lot of usecases, it's a very easy and simple > >> interfaces. > > > > So, you probably late for more than year. Linus W. and others discussed that > > a lot and the points of the choice are listed in documentation IIRC. > > If "deprecated" means there just won't be any new features, but > everything remains as it is, I can live w/ that. But But having to > rewrite lots of applications in the field for the new interface would > be really bad. If you don't want to have new features and OK with broken behaviour, yes, that's fine. > Note that the dev interface is *much* more complex than the sysfs one. > For example, it needs ioctl()s, so this can't be done just w/ a few > lines of shellscript anymore. (which is very common in many embedded > devices) I do it with a shell script, hint: libgpiod is a part of almost all alive Linux distributions. P.S. I don't think I would continue wasting time on the topic, since there is nothing proposed how to improve the case. -- With Best Regards, Andy Shevchenko