David Eccles (gringer)

I've had this same problem. I've attached an example fastq file (2.5kb read) + reference set (12 barcode sequences) that has this problem, as well as another reference sequence set...

And here is another sequence (652bp) that segfaults for me on the same smaller reference sequence set of 4 barcodes, but not with the larger set of 12 barcodes: [graphmap_segfault2.zip](https://github.com/isovic/graphmap/files/612242/graphmap_segfault2.zip)

The segfault for me is in the region_selection code: ``` 0x0000555555676921 in GraphMap::RegionSelectionNoCopy_ (this=0x7fffffffd700, bin_size=217, mapping_data=0x7fffe6eccb80, indexes=..., read=0x555555a50e80, parameters=0x7fffffffd720) at src/graphmap/region_selection.cc:142 142 if (last_update_chromosome[reference_index][position_bin] == (i + 1)) { (gdb)...

curious... lengths of all sequences (4 reverse/forward, so 8 total) in the reference_lengths array is correct, but y_local is not: ``` (gdb) print y_local $24 = 17499632 (gdb) print l_local...

Running through clang++ (with g++ for things that needed omp, because my version of clang can't find omp.h), I got the following error: ``` src/index/index_spaced_hash_fast.cc:844:22: warning: comparison of unsigned expression...

I'm still getting segfaults when running graphmap, mapping ~800Mb of reads to a mitochondrial genome, some reads which have a length larger than the mitochondrial genome length: Version: v0.3.2 Build...

Sorry, still getting a crash when mapping reads to the mitochondrial reference. It might be a different crash, but from my perspective it produces the same end result: ``` $...

I ran a static analysis tool, cppcheck, through the code, which came up with the following warnings / errors: ``` [alignment/alignment_wrappers.cc:108] -> [alignment/alignment_wrappers.cc:87]: (warning) Either the condition 'ret_start_ambiguity!=0' is red...

The first one that I checked (converted_data) could potentially be fixed by replacing uint8_t * with std::vector, but the rabbit hole of templated functions was a bit too deep for...

regarding the index_hash errors, it looks like you've tried using a std::vector to store indexes to a hash table of counts. It would probably work out less painful (for debugging...