ekg / seqwish

alignment to variation graph inducer

MIT License

143 stars 18 forks source link

Consistent segfaults when running seqwish #7

Open apregier opened 5 years ago

apregier commented 5 years ago

I updated to the latest master, but I am still getting a consistent segfault when running. My command is seqwish -a chr22.paf -s chr22.modified.fa.gz -b chr22.graph -d > chr22.gfa 2> chr22.err - I shared my input files with you per e-mail. I am building and running an a vg build docker (v1.10.0-238-g77dc6d6e-t233-build)

ekg commented 5 years ago

What version of GCC did you use to compile seqwish? There was a GCC bug that is AFAICT related to SIMD operations and unaligned memory. Working with the most recent GCC (8) resolved it for me.

ekg commented 5 years ago

6 details this issue.

apregier commented 5 years ago

I just tried with gcc 8.2 and unfortunately I still see the segfault - I expect it is a different problem (maybe data related) since it happens every time rather than sporadically. Did you see my e-mail where I shared input files with you?

ekg commented 5 years ago

Sorry, I still need to work from your test data. I'll loop back to this shortly.

ekg commented 5 years ago

The last commit resolves the case that I was able to reproduce. It seems possible that there are others. This one would have failed sometime during the generation of the GFA output. If you would, please try running your test on the current master.

apregier commented 5 years ago

Seems to have fixed it! I did have to go into bsort and remove the new -lomp linker flag to get it to compile.