On Thu, 1 Apr 2021 15:44:08 +0200 Mauro Carvalho Chehab <mchehab+huawei@xxxxxxxxxx> wrote: > Em Wed, 17 Feb 2021 10:34:34 +0200 > Alexandru Ardelean <alexandru.ardelean@xxxxxxxxxx> escreveu: > > > Since the new change to the IIO buffer infrastructure, the buffer/ and > > scan_elements/ directories have been merged into bufferY/ to have some > > attributes available per-buffer. > > > > This change updates the ABI docs to reflect this change. > > > > The hwfifo attributes are not updated, as for now these should be used > > via the legacy buffer/ directory until they are moved into core. > > > > Signed-off-by: Alexandru Ardelean <alexandru.ardelean@xxxxxxxxxx> > > --- > > Documentation/ABI/testing/sysfs-bus-iio | 85 +++++++++++++++++++++++++ > > 1 file changed, 85 insertions(+) > > > > diff --git a/Documentation/ABI/testing/sysfs-bus-iio b/Documentation/ABI/testing/sysfs-bus-iio > > index d957f5da5c04..f2a72d7fbacb 100644 > > --- a/Documentation/ABI/testing/sysfs-bus-iio > > +++ b/Documentation/ABI/testing/sysfs-bus-iio > > @@ -1118,12 +1118,16 @@ Description: > > > > What: /sys/bus/iio/devices/iio:deviceX/buffer/length > > KernelVersion: 2.6.35 > > +What: /sys/bus/iio/devices/iio:deviceX/bufferY/length > > +KernelVersion: 5.11 > > Contact: linux-iio@xxxxxxxxxxxxxxx > > Description: > > Number of scans contained by the buffer. > > The ABI parser doesn't like things like this: > > $ ./scripts/get_abi.pl validate > > Warning: file Documentation/ABI/testing/sysfs-bus-iio#1167: > What '/sys/bus/iio/devices/iio:deviceX/buffer/length' doesn't have a description > > The main reason is that all properties, including KernelVersion, > Contact and Description are associated to a group of properties. > > To be frank, for me that don't work with IIO, the above ABI > description doesn't sound clear. > > I mean, what's the difference between > /sys/bus/iio/devices/iio:deviceX/buffer/length > and > /sys/bus/iio/devices/iio:deviceX/bufferY/length? > > > If the intention is to tell that: > /sys/bus/iio/devices/iio:deviceX/buffer/length > was obsoleted by: > /sys/bus/iio/devices/iio:deviceX/bufferY/length > > IMO, the right thing would be to move the deprecated definition to > Documentation/ABI/obsolete/ > > If, on the other hand, both are completely identical and > non-obsoleted, why to have both APIs? > > Or did you just missed adding a different description for the > new ABI symbols, but this was dropped due to some merge > conflict? You are right that the version with out the index is obsolete but it's only just become so and there are no realistic way we are going to ever be able to remove it however much we encourage userspace to use the new interface. We are stuck with this bit of backwards compatibility indefinitely. I'm fine with moving it and related definitions over to ABI/obsolete if that makes sense. Jonathan > > Thanks, > Mauro