segfault at end of bcalm (probably due to valgrind error)

[rchikhi]

Problem is in ographBin:

DSK created 32 partitions
[Iterating DSK partitions                ]  0    %   elapsed:   0 min 0  sec   remaining:   0 min 0  sec
Partition 0 has 1518 kmers
Iterated 1518 kmers, among them 350 has leftmin!=rightmin
==18709== Thread 5:
==18709== Conditional jump or move depends on uninitialised value(s)
==18709==    at 0xF7790C: graph4::debruijn() (ographBin.cpp:81)
==18709==    by 0xF7CFA9: bcalm_1::execute()::{lambda(int)#6}::operator()(int) const (bcalm_1.cpp:484)
==18709==    by 0xF84A88: void std::_Bind<bcalm_1::execute()::{lambda(int)#6} (std::_Placeholder<1>)>::__call<void, int&&, 0ul>(std::tuple<int&&>&&, std::_Index_tuple<0ul>) (functional:1264)
==18709==    by 0xF84A33: void std::_Bind<bcalm_1::execute()::{lambda(int)#6} (std::_Placeholder<1>)>::operator()<int, void>(int&&) (functional:1323)
==18709==    by 0xF849E4: std::enable_if<((!std::is_member_pointer<std::_Bind<bcalm_1::execute()::{lambda(int)#6} (std::_Placeholder<1>)> >::value)&&(!std::is_function<std::is_member_pointer>::value))&&(!std::is_member_pointer<std::_Bind<bcalm_1::execute()::{lambda(int)#6} (std::_Placeholder<1>)> ><std::remove_pointer<std::is_member_pointer>::type>::value), std::result_of<std::is_member_pointer& (int&&)>::type>::type std::__invoke<std::_Bind<bcalm_1::execute()::{lambda(int)#6} (std::_Placeholder<1>)>, int>(std::result_of, int&&) (functional:202)
==18709==    by 0xF8498F: std::result_of<bcalm_1::execute()::{lambda(int)#6} (&(int&&))(std::_Placeholder<1>)>::type std::reference_wrapper<std::_Bind<bcalm_1::execute()::{lambda(int)#6} (std::_Placeholder<1>)> >::operator()<int>(int&&) const (functional:435)
==18709==    by 0xF848BF: void std::_Bind_simple<std::reference_wrapper<std::_Bind<bcalm_1::execute()::{lambda(int)#6} (std::_Placeholder<1>)> > (int)>::_M_invoke<0ul>(std::_Index_tuple<0ul>) (functional:1700)
==18709==    by 0xF846C3: std::_Bind_simple<std::reference_wrapper<std::_Bind<bcalm_1::execute()::{lambda(int)#6} (std::_Placeholder<1>)> > (int)>::operator()() (functional:1688)
==18709==    by 0xF84548: std::_Function_handler<void (), std::reference_wrapper<std::_Bind_simple<std::reference_wrapper<std::_Bind<bcalm_1::execute()::{lambda(int)#6} (std::_Placeholder<1>)> > (int)> > >::_M_invoke(std::_Any_data const&) (functional:2069)
==18709==    by 0xFA6780: std::function<void ()>::operator()() const (functional:2439)
==18709==    by 0xFA5A07: std::__future_base::_Task_setter<std::unique_ptr<std::__future_base::_Result<void>, std::__future_base::_Result_base::_Deleter>, void>::operator()() (future:1264)
==18709==    by 0xFA4EAD: std::_Function_handler<std::unique_ptr<std::__future_base::_Result_base, std::__future_base::_Result_base::_Deleter> (), std::__future_base::_Task_setter<std::unique_ptr<std::__future_base::_Result<void>, std::__future_base::_Result_base::_Deleter>, void> >::_M_invoke(std::_Any_data const&) (functional:2025)

[Malfoy]

OgraphBin is still in dev and should not be yet used.

[rchikhi]

so we should use graph3?

[rchikhi]

ograph3 has an issue as well:

[..]
[Iterating DSK partitions                ]  75   %   elapsed:   3 min 17 sec   remaining:   1 min 6  sec
Partition 25 has 1520 kmers
Loaded 503 doubled kmers for partition 25
Iterated 1520 kmers, among them 403 has leftmin!=rightmin
==7688== Conditional jump or move depends on uninitialised value(s)
==7688==    at 0xF8C928: graph3::debruijn() (ograph.cpp:362)
==7688==    by 0xF95771: bcalm_1::execute()::{lambda(int)#6}::operator()(int) const (bcalm_1.cpp:485)
[..]

To reproduce it:

../build/bcalm -in ~/bcalm-data/lambda-k16.fa -k 16 -abundance 1

https://transfer.sh/9h7XC/lambda-k16.fa (link will expire soon)

[Malfoy]

I'll take care of it.
Thanks for the example

[rchikhi]

okay thanks!

[Malfoy]

I can't reproduce the error.
But I see a possible reason for bcalm to crash.
I tried a fix, tell me if it work for you.

[rchikhi]

Indeed, with this new version, I do not get a valgrind error during the execution anymore!
This bug was unrelated then, but thanks for fixing it.

The crash remains at the end of the execution, however it disappears when bcalm is compiled in debug mode. Valgrind output is not helpful without debug information.

==9529== Invalid read of size 4
==9529==    at 0x7091E2: __tcf_3 (in /home/rayan/bcalm-gatb/build/bcalm)
==9529==    by 0x5CCAB28: __run_exit_handlers (exit.c:82)
==9529==    by 0x5CCAB74: exit (exit.c:104)
==9529==    by 0x5CB4B4B: (below main) (libc-start.c:321)
==9529==  Address 0x8f0109e1b0 is not stack'd, malloc'd or (recently) free'd

Neither is gdb:

Program received signal SIGSEGV, Segmentation fault.
0x00000000007091e2 in __tcf_3 ()
(gdb) bt
#0  0x00000000007091e2 in __tcf_3 ()
#1  0x00007ffff6c0ab29 in __run_exit_handlers (status=0, listp=0x7ffff6f785a8 <__exit_funcs>, run_list_atexit=run_list_atexit@entry=true) at exit.c:82
#2  0x00007ffff6c0ab75 in __GI_exit (status=<optimized out>) at exit.c:104
#3  0x00007ffff6bf4b4c in __libc_start_main (main=0x6edf50 <main>, argc=7, argv=0x7fffffffdc18, init=<optimized out>, fini=<optimized out>,
    rtld_fini=<optimized out>, stack_end=0x7fffffffdc08) at libc-start.c:321
#4  0x00000000006ee9cf in _start ()

I'm on it.

[rchikhi]

it was related to a call in initBinSeq, that I just commented out because we don't use binSeq at the moment.