Segmentation Fault - Isoseq3 about pbbioconda

[rsharris]

The details (such as they are) are in this comment in issue 549: #549 (comment)

As far as how to use the binary — download it and run it (giving an explicit path to where it is on your machine). Or put it in some directory that's defined in your $PATH, and run it (just using the binary's file name).

[armintoepfer]

What OS and kernel are you using?

[rsharris]

I have the same problem. Installed from bioconda earlier today. isoseq3 seems to segfault right away regardless of what's the rest of the command line looks like.

I installed this earlier today in a fresh conda environment.
isoseq3-3.8.2 h9ee0642_0 1.6 MB bioconda

Am trying to do perform the steps at https://isoseq.how/clustering/cli-workflow.html , but have been advised to use isoseq3 instead of isoseq. I successfully ran step 2 which uses lima. At step 3 I try isoseq3 refine and get the segfault.

uname -a
Linux fletony 4.19.0-21-amd64 #1 SMP Debian 4.19.249-2 (2022-06-30) x86_64 GNU/Linux

[jamestwebber]

I'm also seeing this again on a new machine, same as back in #549...

Output of uname:
Linux 4.19.0-23-cloud-amd64 #1 SMP Debian 4.19.269-1 (2022-12-20) x86_64 GNU/Linux

[jamestwebber]

The solution from back then worked for me now (downloaded the binaries directly) but it would be really nice if the conda packages actually worked.

[rsharris]

Thanks @jamestwebber . Unfortunately the binary doesn't work for me. I just get "Illegal instruction" instead of a segfault.

[armintoepfer]

Illegal instructions means that you are using an ancient CPU that does not support SSE4.1

[armintoepfer]

@jamestwebber it's strange what conda did to the binary. I'll try to re-upload it...

[rsharris]

@armintoepfer

Illegal instructions means that you are using an ancient CPU that does not support SSE4.1

Inferring from what I see above, the conda package doesn't rebuild isoseq3, it merely loads a binary (am I correct?). In which case even if the conda package's isoseq3 didn't segfault, I'd still have the illegal instruction issue. Right?

As for why I was running on such an old CPU, I was running on the primary machine attached to disk where my sequenced data is. The only restriction I noticed (when I was reading installation instructions) was that I needed my bioconda environment to be python 2.7.

I need to ask, then, if I locate a machine that supports SSE4.1 should that be sufficient?

[armintoepfer]

For Iso-Seq, you don't need python. Just a machine that isn't older than from ~2007.

[rsharris]

@armintoepfer &J.

[philoel]

The solution from back then worked for me now (downloaded the binaries directly) but it would be really nice if the conda packages actually worked.

since I'm getting a Segfault too, could we write here how we can go about downloading and then using the binaries directly? This sounds pretty magical but also mysterious!

Maybe it helps the next person who needs to download binaries too, but doesn't have the experience to know what's involved in this.

[rsharris]

But I would guess when the latest version number changes, you'd have to consider that when you download the binary.

[armintoepfer]

I have uploaded a new version. If that still segfaults, but not the github binary itself, then conda is corrupting the binary.