The CLOCK_REALTIME clock may jump in time, which can be a huge problem
when computing timeouts. Use the CLOCK_MONOTONIC one instead, which does
not have this problem.

Signed-off-by: Paul Cercueil <[email protected]>

When the buffer creation failed, entry->buf was left to whatever error
pointer it had previously, instead of being reset to NULL; cleanup code
would subsequently try to destroy the buffer once again with an invalid
pointer.

Signed-off-by: Paul Cercueil <[email protected]>

IIOD v0.x supports enqueueing cyclic blocks, but it needs to know about
it when the iio_buffer is opened through the "OPEN" command. As
iiod-client calls this command when the buffer is to be enabled, pass an
extra "cyclic" parameter to .buffer_enable(), always set to "false" for
now.

Signed-off-by: Paul Cercueil <[email protected]>

Instead of continuously sending the same block again and again, set the
"cyclic" parameter of the .enable_buffer() backend callback to the value
of the "cyclic" parameter of the last call to iio_buffer_enqueue().

Signed-off-by: Paul Cercueil <[email protected]>

Signed-off-by: Travis F. Collins <[email protected]>

Signed-off-by: Travis F. Collins <[email protected]>

Signed-off-by: Travis F. Collins <[email protected]>

Signed-off-by: Travis F. Collins <[email protected]>

Signed-off-by: Travis F. Collins <[email protected]>

Signed-off-by: Travis F. Collins <[email protected]>

Signed-off-by: Travis F. Collins <[email protected]>