Re: [gpsd-dev] Major progress!

[Greg Troxel]

"Eric S. Raymond" <address@hidden> writes:

> Actually, that's a pretty good clue to what's going on.  There used to be
> two ways a test could terminate:
>
> (1) gpsd recognizes a final "# EOF\n" comment packet injected by gpsfake.
>     It shuts down the device pty.
>
> (2) gpsd waits on select.  The test code notices it has run out of data.  
> After
>     a timeout of CLOSE_DELAY it removes the device.  This is the code path 
> that
>     produced flaky errors.
>
> You've fingered the commit where method 2 was removed.

I will look at this later when I have time.  But I will observe that not
only did gpsd not terminate (it was idle, with reasonable accumulated
cpu usage), but gpsfake was spinning, having gotten to many minutes of
CPU time with no end in sight.
So that may be a separate bug.

pgpuUpdndrUxW.pgp
Description: PGP signature