Closed kinghuang closed 1 year ago
Hey @kinghuang , this is unexpected to me since our more strict dbt-core version is mostly for the dbt-fal CLI, not for the adapter. Curious if this happens for you for dbt-core==1.5.1 and latest dbt-fal version?
We had this working fine for 1.5.1 just recently.
I'm digging into this some more. It looks like I might be encountering two different problems, one for the core dump and one for the version pinning. I'll update this issue once I figure out the details.
I've tracked the core dump to NumPy, specifically when running in a container on Amazon Graviton instances. The issue is covered by numpy/numpy#24028.
dbt-fal indirectly imports numpy, hence why I was experiencing the problem when using a fal profile.
I'll close this issue, and open a new one about the detail in dbt-core dependencies. Thanks!
Thank you for checking!
Describe the bug
I'm encountering a core dump when dbt-fal 1.5.4 is loaded with dbt-core 1.5.2.
dbt-fal 1.5.5 through 1.5.7 has a stated dependency for dbt-core
>= 1.5, <= 1.5.1
.https://github.com/fal-ai/fal/blob/76d9d7502e4e16c0a02ca7387701a0b1cdb2e166/projects/adapter/pyproject.toml#L25
But, dbt-fal 1.5.4 and some earlier versions had a more permissive
>=1.5.0, <1.6
.https://github.com/fal-ai/fal/blob/529c9311dcc90e1c46386b349e0629fe5fdef894/projects/adapter/pyproject.toml#L20
When I'm installing everything along with
dbt-core==1.5.2
ordbt-core==1.5.3
, this results indbt-fal
being resolved to1.5.4
. Is this a valid configuration? What is the reason for the tighter dependency in dbt-fal 1.5.5 through 1.5.7?Your environment
How to reproduce
Expected behavior
With a profile and project that uses fal, be able to run
dbt debug
.Actual behavior
A core dump occurs when dbt loads fal. This can be reproduced to the following snippet.
Screenshots
Additional context