Segmentation fault with parsing of alignment at 'Compressing aligmnent patterns...'

[jgolob]

Greetings! Thank you for the well-supported re-implementation of raxml.

When attempting to parse an alignment, I am consistently running into a segmentation fault with V1.0.2. This has reproduced with the same alignment on multiple systems / setups. --check is able to complete successfully.

The log (with --log debug):

RAxML-NG v. 1.0.2 released on 22.02.2021 by The Exelixis Lab.
Developed by: Alexey M. Kozlov and Alexandros Stamatakis.
Contributors: Diego Darriba, Tomas Flouri, Benoit Morel, Sarah Lutteropp, Ben Bettisworth.
Latest version: https://github.com/amkozlov/raxml-ng
Questions/problems/suggestions? Please visit: https://groups.google.com/forum/#!forum/raxml

System: Intel(R) Core(TM) i7-8086K CPU @ 4.00GHz, 6 cores, 31 GB RAM

RAxML-NG was called at 22-Jul-2021 18:15:07 as follows:

raxml-ng --parse --msa ra.fasta --seed 12345 --model GTR+G --log debug

Analysis options:
  run mode: Alignment parsing and compression
  start tree(s): 
  random seed: 12345
  tip-inner: OFF
  pattern compression: ON
  per-rate scalers: OFF
  site repeats: ON
  branch lengths: proportional (ML estimate, algorithm: NR-FAST)
  SIMD kernels: AVX2
  parallelization: coarse-grained (auto), PTHREADS (auto)

RBA partial loading: OFF
|noname|   |GTR+FO+G4m|   ||
[00:00:00] Reading alignment from file: ra.fasta
Failed to load as IPHYLIP: Unable to parse PHYLIP file: ra.fasta
 (LIBPLL-231): Invalid number of sequences in header
Failed to load as PHYLIP: Unable to parse PHYLIP file: ra.fasta
 (LIBPLL-231): Invalid number of sequences in header
[00:00:01] Loaded alignment with 46991 taxa and 12777 sites
[00:00:01] Extracting partitions... 
[00:00:01] Checking the alignment...

WARNING: Fully undetermined columns found: 60

NOTE: Reduced alignment (with duplicates and gap-only sites/taxa removed) 
NOTE: was saved to: /working/ra.fasta.raxml.reduced.phy
[00:00:47] Compressing alignment patterns... 
Segmentation fault

The alignment is gigantic (~ 500mb) making it a bit impractical to share directly.

[amkozlov]

Hello, thank you for reporting! But i'm afraid it will be difficult to diagnose this problem without having the MSA file. Could you please share it via GDrive etc.?

On a different note: most likely, it will make little sense to run raxml-ng on a dataset with such dimensions (#taxa > #sites) due to insufficient signal. We have discussed this issue quite a lot on the raxml google group, see e.g. these threads: https://groups.google.com/g/raxml/c/13MxNQvve-Y/m/4rLAdj3MAAAJ https://groups.google.com/g/raxml/c/upsTachi-Nc/m/etOL1bW0BAAJ

[amkozlov]

If the issue still exists with the latest raxml-ng version, please reopen and provide input files.