Recently (yesterday) updated from 2022.2 to 2022.11 however on my Linux (Debian Buster) box qiime fails with a hardware error:
[7] 360653 illegal hardware instruction
Googling/tracing etc. suggested that this COULD be due to the Unifrac conda packages were compiled with a particular hardware dependency (GPU perhaps?). So I deleted both unifrac and unifrac-binaries from the install and this fixes the error messaging - but of course unifrac is an important component needed for many of my analyses. I wonder if a) others have experienced this (running a pretty old hardware though on a multiprocessor server from Dell) and if b) there is a simple fix like setting and environment variable like "export UNIFRAC_USE_GPU=N" (which did NOT work when unifrac is installed). Thanks.
Ken
Update:
I uninstalled unifrac (and unifrac-binaries) and then reinstalled from conda repo and I see there is now an "updated" version of the binaries file to bioconda/linux-64::unifrac-binaries-1.2-h73f19ed_1 from unifrac-binaries bioconda/linux-64::unifrac-binaries-1.1.1-h15a0faf_4 and this has completely fixed the program. So thanks to the developers!
Spoke too soon. The problem persists. I see Daniel McDonald encountered a similar problem with the 2022.2 build that was not resolved for that situation. Seems like a compilation quirk on my hardware (standard 2011ish Dell box).
qiime --help
QIIME is caching your current deployment for improved performance. This may take a few moments and should only happen once per deployment.
[1] 3815128 illegal hardware instruction qiime --help
I have uninstalled unifrac unifrac-binaries using conda and conda also uninstalls q2-diversity plugins. Reinstalling unifrac and unifrac-binaries does NOT cause the problem - only trying to add back the q2-diversity plugins causes this issue. If other architectures do not see this, I will just use 2022.2 until a newer 2023 release appears (and perhaps the gods will smile on my architecture again). This is probably a black hole not worth pursuing of little-ole me - the older version seems to work just fine. Or maybe I can spring for a new box to run it on! 10 years is a long time to be using the same old silicon.
Thanks for following up - sorry to hear you're still running into issues!
How are you installing QIIME 2 on your machine? Through conda? If so, can you try running conda clean --all, removing your current 2022.11 environment, and doing a fresh install? If that doesn't work, we can investigate further!
