#293 'Segmentation fault' during 5-consensus

[MaxL]

Hi,

I'm using wgs-8.1 on paired-end Illumina data.
It crashes at 5-consensus on 017:

A sample (start cut-off):

/dg/maxl/P1+P2_Assembly/./JL_P12Celera/5-consensus/consensus.sh 15 > /dev/null 2>&1
/dg/maxl/P1+P2_Assembly/./JL_P12Celera/5-consensus/consensus.sh 16 > /dev/null 2>&1
/dg/maxl/P1+P2_Assembly/./JL_P12Celera/5-consensus/consensus.sh 17 > /dev/null 2>&1
----------------------------------------END CONCURRENT Tue Jan 27 16:08:38 2015 (3880 seconds)
/dg/maxl/P1+P2_Assembly/./JL_P12Celera/5-consensus/JL_P12_017 failed -- no .success.
================================================================================

runCA failed.

---

Stack trace:

at /usr/local/src/wgs_aka_Celera/wgs-8.1/Linux-amd64/bin/runCA line 1501.
main::caFailure('1 unitig consensus jobs failed; remove /dg/maxl/P1+P2_Assembl...', undef) called at /usr/local/src/wgs_aka_Celera/wgs-8.1/Linux-amd64/bin/runCA line 4859
main::postUnitiggerConsensus() called at /usr/local/src/wgs_aka_Celera/wgs-8.1/Linux-amd64/bin/runCA line 6315

---

Failure message:

1 unitig consensus jobs failed; remove /dg/maxl/P1+P2_Assembly/./JL_P12Celera/5-consensus/consensus.sh to try again

It generates a .cns.err file but no .success file.
Looking in the 017.cns.err file:

Computing unitig consensus for b=0 to e=328017
Working on unitig 327297 (0 unitigs and 38203 fragments)
unitig 327297 detected 37129 contains (5402.26x, 96.71%) 1074 dovetail (183.99x, 3.29%)

Failed with 'Segmentation fault'

Backtrace (mangled):

/usr/local/src/wgs_aka_Celera/wgs-8.1/Linux-amd64/bin/utgcns(_Z17AS_UTL_catchCrashiP7siginfoPv+0x27)[0x41e057]
/lib/x86_64-linux-gnu/libpthread.so.0(+0x10340)[0x7f8ee0fd4340]

[Brian Walenz]

Wow, that's a deep unitig!

In consensus.sh, add '-maxcoverage 100' to the utgcns command. This will sample coverage down to 100x. Of course, you can set it to whatever you want. There is a runCA option for this ('cnsMaxCoverage') but it's easier to edit the script at this time.

If that fails, can you recompile the assembler with debug symbols (gmake BUILDDEBUG=1) and/or run with valgrind ("valgrind utgcns ...options from consensus.sh...)?

Can you upload the unitig layout (tigStore -g gkpStore -t tigStore 1 17 -u 327297 -d layout)?

[Brian Walenz]

I should add that using '-V -V -V' will output a lot of diagnostic info, in particular, progress in constructing the consensus sequence which should show where it is failing.