A bug detected by valgrind-3.16.1 in minimap2-2.20_x64-linux ?

[moold]

Hi heng,

I wonder if this is a bug? I have attached the input files:
map.fa.gz
ref.fa.gz

Here is the CMD and log:

valgrind ./minimap2-2.20_x64-linux/minimap2 --sam-hit-only -Y -a -x asm10 -t 10  ref.fa.gz map.fa.gz >/dev/null

==23501== Memcheck, a memory error detector
==23501== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al.
==23501== Using Valgrind-3.16.1 and LibVEX; rerun with -h for copyright info
==23501== Command: ./minimap2-2.20_x64-linux/minimap2 --sam-hit-only -Y -a -x asm10 -t 10 ref.fa.gz map.fa.gz
==23501== 
[M::mm_idx_gen::0.462*2.01] collected minimizers
[M::mm_idx_gen::0.670*1.70] sorted minimizers
[M::main::0.681*1.69] loaded/built the index for 1 target sequence(s)
[M::mm_mapopt_update::0.694*1.67] mid_occ = 50
[M::mm_idx_stat] kmer size: 19; skip: 19; is_hpc: 0; #seq: 1
[M::mm_idx_stat::0.702*1.67] distinct minimizers: 48421 (64.31% are singletons); average occurrences: 1.544; average spacing: 12.041; total length: 900001
==23501== Thread 4:
==23501== Conditional jump or move depends on uninitialised value(s)
==23501==    at 0x4172B3: krmq_update_min_lc_elem (lchain.c:212)
==23501==    by 0x4172B3: krmq_erase_lc_elem (lchain.c:212)
==23501==    by 0x417914: mg_lchain_rmq (lchain.c:291)
==23501==    by 0x40D099: mm_map_frag (map.c:273)
==23501==    by 0x40E0C3: worker_for (map.c:422)
==23501==    by 0x412481: ktf_worker (kthread.c:47)
==23501==    by 0x5356DD4: start_thread (pthread_create.c:307)
==23501==    by 0x5668EAC: clone (clone.S:111)
==23501== 
==23501== Conditional jump or move depends on uninitialised value(s)
==23501==    at 0x4172B3: krmq_update_min_lc_elem (lchain.c:212)
==23501==    by 0x4172B3: krmq_erase_lc_elem (lchain.c:212)
==23501==    by 0x418074: mg_lchain_rmq (lchain.c:282)
==23501==    by 0x40D099: mm_map_frag (map.c:273)
==23501==    by 0x40E0C3: worker_for (map.c:422)
==23501==    by 0x412481: ktf_worker (kthread.c:47)
==23501==    by 0x5356DD4: start_thread (pthread_create.c:307)
==23501==    by 0x5668EAC: clone (clone.S:111)
==23501== 
==23501== Thread 10:
==23501== Conditional jump or move depends on uninitialised value(s)
==23501==    at 0x41F6B8: mm_seed_select (seed.c:56)
==23501==    by 0x41FB50: mm_collect_matches (seed.c:81)
==23501==    by 0x40B072: collect_seed_hits (map.c:179)
==23501==    by 0x40CF85: mm_map_frag (map.c:252)
==23501==    by 0x40E0C3: worker_for (map.c:422)
==23501==    by 0x412481: ktf_worker (kthread.c:47)
==23501==    by 0x5356DD4: start_thread (pthread_create.c:307)
==23501==    by 0x5668EAC: clone (clone.S:111)
==23501== 
[M::worker_pipeline::36.208*1.01] mapped 27 sequences
[M::main] Version: 2.20-r1061
[M::main] CMD: ./minimap2-2.20_x64-linux/minimap2 --sam-hit-only -Y -a -x asm10 -t 10 ref.fa.gz map.fa.gz
[M::main] Real time: 36.289 sec; CPU: 36.768 sec; Peak RSS: 0.619 GB
==23501== 
==23501== HEAP SUMMARY:
==23501==     in use at exit: 0 bytes in 0 blocks
==23501==   total heap usage: 123,045 allocs, 123,045 frees, 1,028,620,143 bytes allocated
==23501== 
==23501== All heap blocks were freed -- no leaks are possible
==23501== 
==23501== Use --track-origins=yes to see where uninitialised values come from
==23501== For lists of detected and suppressed errors, rerun with: -s
==23501== ERROR SUMMARY: 1896 errors from 3 contexts (suppressed: 0 from 0)

[lh3]

Actually in both cases, the conditional jumps from uninitialized values would lead to a deadend and the results are not used downstream. They are both harmless. I still fixed these anyway. Thanks for the report.

[wangyugui]

@lh3

There is a bug in this patch(3f71478: Heng Li: r1063: fixed an uninitialized access (#752)).

With this patch, the memory usage of 'minimap2 -x sr -k15 -w1 -n12 -f500,1500' will become over 110G from 34G.

And this problem is yet not fixed in current master(7025b0b).

Best Regards

[lh3]

I have got a similar bug report. I will look into this later. Thanks.

[lh3]

It is a bug caused by that commit. Please try the latest github HEAD. I believe it has been fixed.