Rust: Bulk model generator #19499
Rust: Bulk model generator #19499
Conversation
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
paldepind
force-pushed
the
rust-bulk-model-generator
branch
from
204eb1c to
68d2961
Compare
paldepind
force-pushed
the
rust-bulk-model-generator
branch
from
68d2961 to
41e76e2
Compare
There was a problem hiding this comment.
Looks great!
This PR commits the output of running the script, but it makes no changes to the quality of the generated models. I hence think that we don't have to inspect the models closely in order to merge this, but instead fix any issues in follow up work when we notice imprecisions or missing models.
Sounds reasonable.
It's a bit odd that taint reach drops for one project in the DCA run, despite only new models being added. Do you have any idea what's going on there?
| ) | ||
| print(f"Completed cloning {name}") | ||
| else: | ||
| print(f"Skipping cloning {name} as it already exists at {target_dir}") |
There was a problem hiding this comment.
Ideally at some point we should handle the case that the specified git_tag has been updated (by fetching the new version).
| else: | ||
| print( | ||
| f"Skipping database creation for {name} as it already exists at {database_dir}" | ||
| ) |
There was a problem hiding this comment.
Again, it would be nice to determine if we should update the database either because the project or codeql version has been changed. This probably isn't a feature we strictly need in the initial version mind you.
This PR adds a Python script for generating models for a list of projects in bulk. The script imports the existing model runner script.
My thinking is that the script can eventually become language-agnostic and shared, but since it's still experimental and subject to change I've put "rust" in the name (
rust_bulk_generate_mad.py) to make it clear that as of now it's only intended for Rust. I'm thinking we can drop therustprefix at some point if/when we become more confident that it's useful and a decent approach.This PR commits the output of running the script, but it makes no changes to the quality of the generated models. I hence think that we don't have to inspect the models closely in order to merge this, but instead fix any issues in follow up work when we notice imprecisions or missing models.