Re: [gpsd-dev] recent regression failures

[Eric S. Raymond]

Greg Troxel <address@hidden>:
> 
> I updated, and get some regression faiilures.
> 
> At
> 
> commit 580743e54163b3a5bb0f2e049fc92b2da86f60ff
> Author: Gary E. Miller <address@hidden>
> Date:   Tue Nov 5 11:30:59 2013 -0800
> 
> I got
> 
> Regression test FAILED: 7 errors in 86 tests total (0 not found).
> The following test Failed:
> ================================================================
> "test/daemon/naujoks-ntrip.log"
> "test/daemon/naujoks-rtcm2.log"
> "test/daemon/nl402u.log"
> "test/daemon/nl551e.log"
> "test/daemon/rtcm2.log"
> "test/daemon/tcp-test.log"
> "test/daemon/tcp-torture.log"
> ================================================================
> 
> At
> 
> commit 65d03ef7cb7059b84834e36d3425b372cec3b0d6
> Author: Eric S. Raymond <address@hidden>
> Date:   Tue Nov 5 11:02:44 2013 -0500
> 
> I got
> 
> Regression test FAILED: 5 errors in 86 tests total (0 not found).
> The following test Failed:
> ================================================================
> "test/daemon/naujoks-ntrip.log"
> "test/daemon/naujoks-rtcm2.log"
> "test/daemon/rtcm2.log"
> "test/daemon/tcp-test.log"
> "test/daemon/tcp-torture.log"
> ================================================================
> 
> it might have been
> 6c69f66
> that was the last good time I tried; I'm not really sure.

I suspect that if you drop back to here

commit 1ec0312d27a0bc04d1fbef89177a694828afd670
Author: Eric S. Raymond <address@hidden>
Date:   Sat Nov 2 17:42:38 2013 -0400

    GPSD is fully up on Mac OS 10.7 - regression tests pass, PPS is live.

    Greg Troxel noted some previously undocumented Python dependencies of
    xgps, now added to build.txt.

all will be OK; you can bisect in on the causes for individual files
failing. 

naujoks-ntrip.log naujoks-rtcm2.log rtcm2.log:
    These failures probably have to do with endianness detection screwing
    up in the RTCM2 driver. Likely they'll go away now that the magic macro
    from StackOverflow is in place

tcp-test.log tcp-torture.log:
    These are recently-introduced tests for handling of TCP data sources.
    They could be tickling some sort of BSD port bug.

nl402u.log nl551e.log:
    I don't know what's up with these, they're plain vanilla NMEA.
    A diff showing how they're different from expected would be
    useful.

> (BTW, it would be nice if 'scons check' did a "git log --oneline -1" if
> a .git directory existed.  Then saved OUT.check.NNN files would be
> obviously tied to the build version.) 

I'll look into it.
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>

signature.asc
Description: Digital signature