Closed ctb closed 5 years ago
Oddly enough my shorter automated workflow https://github.com/ctb/2018-bcalm-debug doesn't detect any problems. Must not trigger the right problem.
Note that when I compile bcalm from scratch on my HPC the spacegraphcats-twofoo-example above runs just fine and passes the reciprocal edges test.
Thanks for the detailed report! I confirm that I can reproduce the behavior of https://github.com/GATB/bcalm/issues/37#issue-393778335 on my own Linux, when using bioconda's bcalm. The problem doesn't occur when using the current master branch of bcalm. So something must be going on with bioconda's version, I'm on it.
Whew! Finally :). I hate Heisenbugs!
-- Titus Brown, ctbrown@ucdavis.edumailto:ctbrown@ucdavis.edu
On Dec 23, 2018, at 3:09 PM, Rayan Chikhi notifications@github.com<mailto:notifications@github.com> wrote:
Thanks for the detailed report! I confirm that I can reproduce the behavior of #37 (comment)https://github.com/GATB/bcalm/issues/37#issue-393778335 on my own Linux, when using bioconda's bcalm. The problem doesn't occur when using the current master branch of bcalm. So something must be going on with bioconda's version, I'm on it.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/GATB/bcalm/issues/37#issuecomment-449668832, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AADHSEvd6o4r_Mv6i9F83vQlxh3SPe4hks5u8A0_gaJpZM4Zf-k_.
I think I see what's going on. Two bugs: 1) I thought BCALM outputted consecutive unitig IDs, but apparently with some compilers it had some rare inversions (e.g. 9,10,12,11,13,14,..). Likely a multi-threading bug. 2) The procedure used to link unitigs silently and wrongly assumed that unitigs needed to have consecutive IDs. I've commited a fix for both bugs and will release BCALM 2.2.1. @natir could you please update Conda when you have time, or show me how?
you rock! thanks, @rchikhi!
ping! I'd suggest withdrawing bcalm from bioconda until someone has time to fix it.
pinging @natir: bcalm 2.2.1 is released and needs to be updated in bioconda please :)
@epruesse update bioconda bcalm to 2.2.1 in this pull request bioconda/bioconda-recipes#12811
merci!
wonderful!
In #34, I reported that in not-very-reproducible circumstances on my HPC, when I compiled bcalm from source, it produced bad cDBG files. I ended up automating the heck out of the check and being unable to reproduce it so I gave up. But I cleverly never removed the validation code from the spacegraphcats application :)
Fast forward to today, and I found the same problem as in #34 when I installed bcalm using bioconda! Hopefully this time it will be more reproducible. I reproduce the bug on both my HPC and Amazon's Linux box, which kind of makes sense because bioconda provides a binary install so I would hope that it would fail the same way on both :).
To reproduce using spacegraphcats
this is running on Amazon Linux 2 AMI (HVM), SSD Volume Type - ami-01bbe152bf19d0289, region us-west-2. I created a t2-large with 20 GB of disk space, logged in as
ec2-user
, and then ran the following commands (some require answering interactive prompts, sorry):Now, activate and try running the spacegraphcats example:
this fails for me after a few minutes of running with:
Here's what I have installed for bcalm:
cc @natir viz #32.