Installation GCC version + segfaults

[srisarya]

Hi team,

Firstly, thanks for this tool, it looks great and I'm excited to use it.

I'm having some installation issues, I hope you can help with these.

I cloned the repo, entered it, and installed as per the instructions (i.e., ./build.sh, whether I run this script or do the commands manually). The dicciculties I'm having are as below:

• If I install with GCC 8.2.0 (my HPC's own module) or GCC 9.4.0 (in a conda env I set up for installation purposes), RATTLE compiles while printing many warnings. Then says I have missing GLIBCXX_3.4.26, GLIBCXX_3.4.29, and CXXABI_1.3.13.
  • If I then load in GCC 10+ (any version, I tested up to v12, even though this is against the docs), RATTLE runs nicely for rattle cluster, rattle cluster_summary, and rattle cluster_extract, until rattle correct, at which it prints a segmentation fault.
• Alternatively, if I compile from the beginning with GCC 10+ (against the docs, I know), it compiles with many warnings and then fails at rattle correct, printing illegal instruction at 22% of the run for one fastq file.

This occurs with the toy dataset and with my own data (cDNA-PCR nanopore reads, cleaned and adapter trimmed).

I'd assume something is an issue with the HPC I'm working on (some library C++ thing I am not aware of, in which case I'd fix it with installing libraries), but then why does rattle run until error-correcting?

Please advise. Have I missed something crucial? Thanks

[EduEyras]

Hi RATTLE uses the SPOA library in the correction step. It could be that something with that library does not work well in your environment. It could be that it runs out of memory in the correction step. This could happen if the clusters are too large. E.

[srisarya]

Thank you! Turns out it was errors my side, we've fixed it by making a singularity image :) On a related note, is there a plan to make a conda package for RATTLE? It would be really nice, and helpful for workflow management.