Closed spolloni closed 1 year ago
I'm not sure, I haven't seen this behavior and I do use S3 heavily with Entwine. Can you maybe send me a key pair with read access to this data via email? Otherwise I'm not sure how to reproduce this.
just sent those over. By the way, I managed to avoid this problem and successfully scan all the files using the entwine docker container.
Was your failing build using Conda, or was it self built?
using Conda. probably should have mentioned this earlier.
I don't believe this is an issue anymore, a new ticket can be opened with a reproduction case if so.
In attempting to build a large amount (~40K) of laz files, I am encountering segmentation faults at varying moments during the scanning step. The exact command I'm using is:
where
tiles_3443.txt
is a list of S3 file keys. The build is run from a Linux (18.04.3 LTS Bionic Beaver) machine in AWS EC2. After looking at #218, I proceeded to splittiles_3443.txt
into successively smaller subsets in hope of finding an offending corrupt file.Strangely, it seems the segfaults are not deterministic ( I wonder if this has to do with the read from S3?). Out of the ~40K files I am trying to build, I identified a handful that will trigger the segfault on occasion even when built individually. For example:
here is a traceback obtained with
gdb
against the core dump:any clue on what might be causing this?