Re: [gpsd-dev] [PATCH] Fixes various flakiness in PPS handling.

[Fred Wright]

On Wed, 21 Sep 2016, Fred Wright wrote:

[...]

It looks like this one fell through a crack.

I strongly suspect that this bug is also the only reason that
inner_context as an auto variable didn't work, and that making it static
only reduced the bug window without actually fixing the real bug.  It
should work as an auto with this fix, though I didn't bother doing that,
since I wanted to minimize the change in code where I only have limited
testing ability, and making pps_thread_activate() thread-safe isn't
terribly important, anyway (except to get rid of the FIXME :-)).

BTW, was that ugly MAGIC_HAT kludge added just to save a few characters on
the command line, or because specifying the PPS source explicitly on the
command line wasn't working reliably?  If the latter, it was almost
certainly due to this bug, and the kludge should now be unnecessary.

Note in particular that the MAGIC_HAT/MAGIC_LINK hack is not only kludgy,
but also a complete disaster when used with a kernel that's built with
CONFIG_PPS_CLIENT_KTIMER=y, since the latter causes /dev/pps0 to be the
fake software-generated PPS source, which doesn't even run at quite 1Hz.

Fred Wright