BUG: SIGNAL SEGMENTATION FAULT in v1.6.8

[Trevor Bedford]

Hello,

I've been successfully running IQTREE to maintain much of nextstrain.org. It's a fantastic program and I truly appreciate all the effort that's gone into its construction. Most trees I have run without issue. However, I'm getting some consistent SIGNAL SEGMENTATION FAULT for recent Zika virus sequences.

I'm running IQTREE v1.6.8:

Fenrir:zika trvrb$ iqtree -ninit 2 -n 2 -me 0.05 -nt 1 -s iqtree_broken_full.fasta -m GTR

IQ-TREE multicore version 1.6.8 for Mac OS X 64-bit built Nov 19 2018

Developed by Bui Quang Minh, Nguyen Lam Tung, Olga Chernomor,

Heiko Schmidt, Dominik Schrempf, Michael Woodhams.

Host:    Fenrir (AVX, 16 GB RAM)

Command: iqtree -ninit 2 -n 2 -me 0.05 -nt 1 -s iqtree_broken_full.fasta -m GTR

Seed:    880674 (Using SPRNG - Scalable Parallel Random Number Generator)

Time:    Mon Nov 19 18:09:50 2018

Kernel:  AVX - 1 threads (8 CPU cores detected)

And getting the following crash:

Estimate model parameters (epsilon = 0.500)

1. Initial log-likelihood: -29612.371

ERROR: STACK TRACE FOR DEBUGGING:

ERROR: 

ERROR: *** IQ-TREE CRASHES WITH SIGNAL SEGMENTATION FAULT

ERROR: *** For bug report please send to developers:

ERROR: ***    Log file: iqtree_broken_full.fasta.log

ERROR: ***    Alignment files (if possible)

Segmentation fault: 11

-----------------------------------

This is on a alignment of 228 taxa with 10,769 sites. Interesting, if I break the alignment in half and run the same command on either half individually, it works fine. So, there's some interaction effect rather than a specific bad sequence.

The full "broken" alignment is available here: https://s3.amazonaws.com/trvrb/iqtree_broken_full.fasta

The first half of the alignment that's working is available here: https://s3.amazonaws.com/trvrb/iqtree_working_split_1.fasta

The second half of the alignment that's working is available here: https://s3.amazonaws.com/trvrb/iqtree_working_split_2.fasta

In the first case, IQTREE is getting called with: iqtree -ninit 2 -n 2 -me 0.05 -nt 1 -s iqtree_broken_full.fasta -m GTR

The other calls just have the FASTA swapped.

Any help here would be much appreciated.

Thank you,

 - Trevor Bedford

[Trevor Bedford]

I've now determined that this issue is present on a different computer running version 1.6.8, so it is not hardware specific.

Additionally, running with IQTREE version 1.6.6 does not show the issue. I also checked 1.6.7.2 and this doesn't show the issue either. Appears to have arisen in 1.6.8.

This gives me a workaround to stick with 1.6.7.2 for the moment.

Best,

 - Trevor

[Minh Bui]

Hi Trevor, Thanks a lot for localising this issue! Apparently it was introduced in v1.6.8. I’ll have a closer look in 2 weeks or so. But that’s OK as you can still use the 2nd latest version.

Cheers

Minh

--
You received this message because you are subscribed to the Google Groups "IQ-TREE" group.
To unsubscribe from this group and stop receiving emails from it, send an email to iqtree+un...@googlegroups.com.
To post to this group, send email to iqt...@googlegroups.com.
Visit this group at https://groups.google.com/group/iqtree.
For more options, visit https://groups.google.com/d/optout.

[Minh Bui]

Hi Trevor,

I've tested this run on my Mac and Linux and nothing crashed at all. However, I also noticed that this crash occurs when using iqtree package provided by Homebrew for Mac. Did you use this one? If yes, can you please use the official releases provided here:

https://github.com/Cibiv/IQ-TREE/releases/tag/v1.6.8

Unfortunately, I don’t know the homebrew iqtree package, so you may contact them directly. 

Cheers,

Minh