[RFC] inotify implementation

[Richard Hughes]

On Tue, 2007-01-16 at 13:29 +0300, Sergey Lapin wrote:
> Richard Hughes wrote:
> > On Tue, 2007-01-16 at 13:01 +0300, Sergey Lapin wrote:
> >> As I understand inotify API first appeared in glibc-2.4.
> >> As it is well known, many ARM machines still use glibc-2.3.5
> >> (probably for good reason), so have no such implementation.
> > 
> > Would ARM machines running 2.3.5 want to be running the latest HAL?
> Well, since it is better optimized for embedded devices, yes.

Okay, didn't know this.

> > 
> > Also, I think it's safe to just look for sys/inotify.h, as if the user
> > has installed that file, but not added support in the kernel then it is
> > his fault!
> Yeah, but runtime check of init routine would do no harm either, IMHO.

Sure, agree.

> > My personal view is also that we shouldn't support dnotify, just like we
> > don't support DBUS < 0.70 or glib 2.6 - i.e. better versions are in
> > common usage.
> Yeah, but since kernel supports that, and linc doesn't
> so it is dilemma - we add routines ourself or pretend as
> it doesn't exist. As I show in previous email, it is not too big
> amount of code, but I'm curious about if it will be hard to
> maintain, and about direct kernel access ugliness so what is bad -
> support quite a lot of embedded devices in ugly way (while most of
> them have fixed configuration and will probably never change FDI
> files, and that's nice but not that important feature for them),
> or cleanly refuse to do so.

Direct kernel access is very icky if you ask me.

Richard.