Rust: Update legacy MaD models 2 #19942
Conversation
geoffw0
added
no-change-note-required
|
Accepted test regressions. There are almost no results left for We also lose some results in a test case involving a loop, similar to the one discussed in #19934 (comment) . |
| - ["repo:https://github.com/RustCrypto/traits:digest", "<_ as crate::digest::Digest>::chain_update", "Argument[0]", "hasher-input", "manual"] | ||
| - ["repo:https://github.com/RustCrypto/traits:digest", "<_ as crate::digest::Digest>::digest", "Argument[0]", "hasher-input", "manual"] | ||
| - ["repo:https://github.com/stainless-steel/md5:md5", "crate::compute", "Argument[0]", "hasher-input", "manual"] | ||
| - ["<digest::core_api::wrapper::CoreWrapper as digest::Update>::new_with_prefix", "Argument[0]", "hasher-input", "manual"] |
There was a problem hiding this comment.
This looks a bit strange. The digest::Update trait does not have a new_with_prefix as far as I know. I think it doesn't have chain_update or digest either.
There was a problem hiding this comment.
I'll investigate this...
There was a problem hiding this comment.
In real world (MRVA) projects we mostly don't get a canonical path at all on new_with_prefix, when we do it's something like the above or <_ as digest::digest::Digest>::new_with_prefix. The latter looks better actually, so I've switched to that.
I plan to convert this to a trait model at some point, which will be more robust, but I don't think we (quite) have everything in place for that yet.
| - ["<_ as tokio::io::util::async_read_ext::AsyncReadExt>:::read_f32_le", "Argument[self]", "ReturnValue.Future.Field[core::result::Result::Ok(0)]", "taint", "manual"] | ||
| - ["<_ as tokio::io::util::async_read_ext::AsyncReadExt>::read_f64", "Argument[self]", "ReturnValue.Future.Field[core::result::Result::Ok(0)]", "taint", "manual"] | ||
| - ["<_ as tokio::io::util::async_read_ext::AsyncReadExt>::read_f64_le", "Argument[self]", "ReturnValue.Future.Field[core::result::Result::Ok(0)]", "taint", "manual"] | ||
| - ["<_ as core::iter::traits::iterator::Iterator>::chain", "Argument[self]", "ReturnValue", "taint", "manual"] |
There was a problem hiding this comment.
These probably do not belong in the tokio models anymore.
There was a problem hiding this comment.
I agree, but I'm keen to save moving stuff around for after these four PRs are merged (to prevent merge conflicts and confusing diffs).
I do have an issue tracking this.
Co-authored-by: Arthur Baars <aibaars@github.com>
rust/ql/lib/codeql/rust/frameworks/rustcrypto/rustcrypto.model.yml
Outdated
Show resolved
Hide resolved
| @@ -11,8 +11,8 @@ fn test_hash_algorithms( | |||
|
|
|||
| // MD5 | |||
| _ = md5::Md5::digest(harmless); | |||
| _ = md5::Md5::digest(credit_card_no); // $ Alert[rust/weak-sensitive-data-hashing] | |||
| _ = md5::Md5::digest(password); // $ Alert[rust/weak-sensitive-data-hashing] | |||
| _ = md5::Md5::digest(credit_card_no); // $ MISSING: Alert[rust/weak-sensitive-data-hashing] | |||
There was a problem hiding this comment.
Currently getStaticTarget() fails on md5::Md5::digest. This are the relevant lines:
pub type Md5 = CoreWrapper<Md5Core>;
impl<D: FixedOutput + Default + Update + HashMarker> Digest for D {
fn digest(data: impl AsRef<[u8]>) -> Output<Self> { ... }
}There was a problem hiding this comment.
@hvitved I guess the type variable in for D is the cause of the trouble, isn't it?
| let id: i32 = row.get("id"); // $ MISSING: database-read | ||
| let name: &str = row.try_get("name")?; // $ MISSING: database-read | ||
| let age: i32 = row.try_get("age").unwrap(); // $ MISSING: database-read |
There was a problem hiding this comment.
The problem seems to be that the type inferencer fails to infer the type of row. The following trick should make the tests pass again. We ought to fix the type inferencer, but the purpose of this test case is to validate whether the postgres models work. @hvitved Any idea why the type inferencer didn't work here?
| let id: i32 = row.get("id"); // $ MISSING: database-read | |
| let name: &str = row.try_get("name")?; // $ MISSING: database-read | |
| let age: i32 = row.try_get("age").unwrap(); // $ MISSING: database-read | |
| let row: &postgres::Row = row; | |
| let id: i32 = row.get("id"); // $ database-read | |
| let name: &str = row.try_get("name")?; // $ database-read | |
| let age: i32 = row.try_get("age").unwrap(); // $ database-read |
There was a problem hiding this comment.
I've accepted this, and we can now see that the model works. But I've also added back the original test, as I don't like the idea we could be gradually simplifying our tests and hiding problematic areas.
….yml Co-authored-by: Arthur Baars <aibaars@github.com>
Co-authored-by: Arthur Baars <aibaars@github.com>
Update some more legacy MaD models to the new model format (continues from #19934 , but should be independent of that).