Segmentation fault

[sjackman]

Any thoughts on why bfc segfaulted here? It had been running for 11.5 hours and was using 952 GB of RAM when it crashed.

❯❯❯ bfc -v
r181
❯❯❯ gunzip -c data.fq.gz | bfc -t16 -s14G -d data.bfc /dev/stdin /dev/null
[M::main] applied `-k 35 -b 37'
[M::bfc_count_cb] read 663715 sequences
[M::bfc_count_cb] read 663537 sequences
[M::bfc_count_cb @248.2*112.3%] processed 663715 sequences; # distinct k-mers: 4945319
[M::bfc_count_cb] read 663527 sequences
…
[M::bfc_count_cb] read 663411 sequences
[M::bfc_count_cb @39980.6*1512.8%] processed 663497 sequences; # distinct k-mers: 12783060251
[M::bfc_count_cb] read 663383 sequences
[M::bfc_count_cb @39987.7*1512.8%] processed 663411 sequences; # distinct k-mers: 12784917041
[M::bfc_count_cb] read 663362 sequences
Segmentation fault (core dumped)

[aclum]

Was there any resolution to this? I'm having a similar issue with a similar amount of distinct kmers.

[sjackman]

Not from me

[aclum]

When I run valgrind I get

==9907== Thread 4:
==9907== Invalid read of size 8
==9907== at 0x402F98: bfc_ch_insert (htab.c:75)
==9907== by 0x404BF1: bfc_kmer_insert (count.c:61)
==9907== by 0x404BF1: worker_count (count.c:86)
==9907== by 0x401B22: ktf_worker (kthread.c:41)
==9907== by 0x5355743: start_thread (in /os/lib64/libpthread-2.22.so)
==9907== Address 0x35c8f7ec0 is 0 bytes after a block of size 32,768 alloc'd
==9907== at 0x4C2B46E: realloc (in /os/usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
==9907== by 0x403501: kh_resize_cnt (htab.c:9)
==9907== by 0x403501: kh_put_cnt (htab.c:9)
==9907== by 0x403501: bfc_ch_insert (htab.c:73)
==9907== by 0x404BF1: bfc_kmer_insert (count.c:61)
==9907== by 0x404BF1: worker_count (count.c:86)
==9907== by 0x401B22: ktf_worker (kthread.c:41)
==9907== by 0x5355743: start_thread (in /os/lib64/libpthread-2.22.so)
==9907==
==9907== Thread 3:
==9907== Invalid write of size 8
==9907== at 0x402731: bseq_read (bseq.c:65)
==9907== by 0x40470C: bfc_count_cb (count.c:97)
==9907== by 0x401C5C: ktp_worker (kthread.c:106)
==9907== by 0x5355743: start_thread (in /os/lib64/libpthread-2.22.so)
==9907== Address 0xc010 is not stack'd, malloc'd or (recently) free'd
==9907==
==9907==
==9907== Process terminating with default action of signal 11 (SIGSEGV)
==9907== Access not within mapped region at address 0xC010
==9907== at 0x402731: bseq_read (bseq.c:65)
==9907== by 0x40470C: bfc_count_cb (count.c:97)
==9907== by 0x401C5C: ktp_worker (kthread.c:106)
==9907== by 0x5355743: start_thread (in /os/lib64/libpthread-2.22.so)
==9907== If you believe this happened as a result of a stack
==9907== overflow in your program's main thread (unlikely but
==9907== possible), you can try to increase the size of the
==9907== main thread stack using the --main-stacksize= flag.
==9907== The main thread stack size used in this run was 8388608.
==9907==
==9907== HEAP SUMMARY:
==9907== in use at exit: 48,324,538,206 bytes in 5,370,731 blocks
==9907== total heap usage: 175,663,029 allocs, 170,292,298 frees, 120,721,831,998 bytes allocated
==9907==

[GuillaumeHolley]

Just had a Segmentation Fault (SIGSEGV) when bfc_correct started (right after bfc_count_cb finished).

[myronpeto]

I was getting a segmentation fault at that exact location but after specifying the approximate genome size and giving it more threads it completed. I also noticed that it automatically set -k 33 and -b 37. Unsure which, if any, of those changes got it working.