peregrine segfaults during tracking about libswiftnav-legacy HOT 6 CLOSED

Thanks for this detailed bug report.

Is there any chance you could upload your 1/10 size data set somewhere so I can try to reproduce the segfault?

On 27 Jun 2014, at 00:42, Niels Laukens [email protected] wrote:

I'm trying to use peregrine to analyze a fairly large file (~12 minutes of samples, 5.6GiB piksi format). Acquisition runs fine, but it segfaults systematically at 19% tracking (with or without re-using the acquisition data):

$ peregrine -a samples.piksi
2014-06-27 08:23:45,959 [INFO] root: Skipping acquisition, loading saved acquisition results.
2014-06-27 08:24:43,866 [INFO] peregrine.tracking: Tracking starting
2014-06-27 08:24:43,871 [DEBUG] peregrine.tracking: Tracking 5 channels, PRNs [15, 12, 17, 24, 18]
Tracking (CH: 1/5) 19% ETA: 0:03:00 |######### |
Segmentation fault
From the backtrace however, it seems like the problem is inside libswiftnav, that's why I report it here:

#0 0x00007fae52fae893 in track_correlate (
samples=0x7fac82990c28 ""...,
code=0xfabf20 ""..., init_code_phase=0x7fff277ea0b0, code_step=0.062499877161046859, init_carr_phase=0x7fff277ea0c0, carr_step=1.5700294860690449,
I_E=0x7fff277ea0d0, Q_E=0x7fff277ea0e0, I_P=I_P@entry=0x7fff277ea0f0, Q_P=Q_P@entry=0x7fff277ea100, I_L=I_L@entry=0x7fff277ea110, Q_L=Q_L@entry=0x7fff277ea130,
num_samples=num_samples@entry=0x7fff277ea0a0) at /usr/local/src/libswiftnav/src/correlate.c:57
#1 0x00007fae5297e81a in __pyx_pf_8swiftnav_9correlate_2track_correlate (__pyx_self=, __pyx_v_settings=0x7fae538668c0, __pyx_v_caCode=,
__pyx_v_remCarrPhase=, __pyx_v_carrFreq=, __pyx_v_remCodePhase=, __pyx_v_codeFreq=, __pyx_v_rawSignal=)
at swiftnav/correlate.c:2314
#2 __pyx_pw_8swiftnav_9correlate_3track_correlate (__pyx_self=, __pyx_args=, __pyx_kwds=) at swiftnav/correlate.c:2153
#3 0x000000000052c6d5 in PyEval_EvalFrameEx ()
#4 0x000000000055c594 in PyEval_EvalCodeEx ()
#5 0x000000000052ca8d in PyEval_EvalFrameEx ()
#6 0x000000000052cf32 in PyEval_EvalFrameEx ()
#7 0x000000000055c594 in PyEval_EvalCodeEx ()
#8 0x00000000005b7392 in PyEval_EvalCode ()
#9 0x0000000000469663 in ?? ()
#10 0x00000000004699e3 in PyRun_FileExFlags ()
#11 0x0000000000469f1c in PyRun_SimpleFileExFlags ()
#12 0x000000000046ab81 in Py_Main ()
#13 0x00007fae58dedec5 in __libc_start_main (main=0x46ac3f

, argc=4, argv=0x7fff277eaab8, init=, fini=, rtld_fini=, stack_end=0x7fff277eaaa8)
at libc-start.c:287
#14 0x000000000057497e in _start ()
The problem does not seem to be correlated with the length of the file (I reduced the file to one tenth of its length, and peregrine still segfaults at 19%), so I'm guessing it segfaults when trying to track the second satellite.

Some system information that may be relevant:

swift-nav/libswiftnav@47f3827
swift-nav/libswiftnav-python@8143b5c
swift-nav/peregrine@e67d0c4
gcc version 4.8.2 (Ubuntu 4.8.2-19ubuntu1)
libfftw3 3.3.3-7ubuntu3
Python 2.7.6
Ubuntu 14.04
x86_64 / amd64
—
Reply to this email directly or view it on GitHub.

from libswiftnav-legacy.

Is there any chance you could upload your 1/10 size data set somewhere so I can try to reproduce the segfault?

Yes. I've reduced it to around 1/20th size (still 300MB), and it still segfaults. I've uploaded it (temporarily) here:
ftp://user1_ro:[email protected]/user1/gps-sample.piksi

I do want to confess that the file is not an original capture. For storage reasons, I converted the original Piksi capture to 1-bit-format (since, from what I understood from the code, the MAX2769 is set to 1bit mode), and back to piksi format. Since acquisition runs fine, I don't think I missed something by doing so, but I wanted to provide you with all info.

from libswiftnav-legacy.

Hi Niels,

Sorry this took a while. I just pushed a patch to peregrine master (swift-nav/peregrine@12dfe7a7) that fixes the segfault. Just 4 characters missing :) Let me know if it works on your end.

Thanks,
Fergus

from libswiftnav-legacy.

Unfortunately, I still get a segfault, seemingly at the same place: at the end of tracking the first sat, or at the beginning of tracking the second; segfault in track_correlate() in correlate.c:57. (Using swift-nav/peregrine@87fc454)

from libswiftnav-legacy.

Minor update: I've updated to swift-nav/libswiftnav@621b680 and swift-nav/libswiftnav-python@6ff451b, but the problem persists.

from libswiftnav-legacy.

Ping?

from libswiftnav-legacy.