Re: [gpsd-dev] gpsd crash

[Eric S. Raymond]

address@hidden <address@hidden>:
> Hello,
> 
> i habe done some investigations on the crah i observed after the awake of my 
> netbook from sleep.
> 
> 1. gpsd crash when i exit gpsmon.
> 2. gpsd crash in line 2172 of gpsd.c
> 3. Here is the gdb output of
> 
> p *device 
> 
> $3 = {gpsdata = {set = 33554434, online = 0, gps_fd = -1, fix = {
>       time = 1415333875, mode = 3, ept = 0.0050000000000000001, 
>       latitude = 52.364035350068853, epy = 12.773699940785612, 
>       longitude = 9.6584529921558975, epx = 31.739733263843473, 
>       altitude = 63.404462945251616, epv = 37.66646278533706, 
>       track = 84.403599220853238, epd = nan(0x8000000000000), 
>       speed = 0.53741792350378159, eps = 63.479466527686945, 
>       climb = -0.19825986859897671, epc = 75.332925570674121}, 
>     separation = 46.367540598185151, status = 1, satellites_used = 6, dop = {
>       xdop = 2.1159822175895648, ydop = 0.85157999605237411, 
>       pdop = 2.8079422680479702, hdop = 1.2, vdop = 1.6376722950146547, 
>       tdop = 1.7064308073137007, gdop = 3.2857945585260908}, 
>     epe = nan(0x8000000000000), skyview_time = 1415333872.99, 
>     satellites_visible = 9, skyview = {{ss = 39, used = true, PRN = 12, 
> 
> Note that gpsfd is -1
> 
> Perhaps this helps hunting this bug.
> 
> Reinhard

I think it does.

I suspect your crash is in this section of code:

    while (0 == signalled) {
        fd_set efds;
        switch(gpsd_await_data(&rfds, &efds, maxfd, &all_fds, &context.errout))
        {
        case AWAIT_GOT_INPUT:
            break;
        case AWAIT_NOT_READY:
            for (device = devices; device < devices + MAXDEVICES; device++)
                if (allocated_device(device) && 
FD_ISSET(device->gpsdata.gps_fd, &efds)) {
                    deactivate_device(device);
                    free_device(device);
                }
            continue;
        case AWAIT_FAILED:
            exit(EXIT_FAILURE);
        }

Most likely what's happening is that on receipt of wakeup signal it is
going to the AWAIT_NOT_READY case and dying somewhere in the
deactivate() or free_device() code.

I think I fixed a bug here around the end of the year, so it might be your
problem is solved already; please check with head.  If not, the first thing 
to try is commenting out the deactivate() and free_device() calls to see if
that prevents the crash.
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>