-
Notifications
You must be signed in to change notification settings - Fork 1.6k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Rust: More tests for rust/deadcode #17923
Conversation
} | ||
|
||
const _: () = { | ||
_ = 1; // $ SPURIOUS: Alert[rust/dead-code] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is a child of a Const (Const
> BlockExpr
> StmtList
> ExprStmt
> AssignmentExpr
); I'm not sure if we want to exclude these from the query, or add them to the CFG so that they're considered reachable.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
One argument for including them in the CFG is that const
code could contain unreachable code. If we exclude constant things, then we will not be able to detect this and other things (unused variables, etc.). An argument in the other direction is that code evaluated at compile time is unlikely to contain security issues. Or maybe it can?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It feels like const
code is less likely to contain security issues, but still could. Certainly for data flow - for example the value of a security relevant constant (e.g. enabling or disabling a feature) could be computed in const
code.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Do you know what steps we need to take to include them in the CFG? Is it just a case of adding the entry points somewhere???
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I've created an issue to track this, but I could do with some pointers (or someone else to take on the issue, if that is easier).
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks! I'll be looking at that together as well as the async
problem 👍
if cond() { | ||
do_something(); | ||
bail_1!(); // $ SPURIOUS: Alert[rust/dead-code] | ||
do_something(); // $ MISSING: Alert[rust/dead-code] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is an issue with the workarounds for PostOrderTree
and macro nodes. I'll figure out a way to fix it.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I've created an issue to track this, I know what to do but I have other priorities.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Examples look good to me 👍
Do you know why the variables used with await
are spuriously marked unused? If not I can look into that.
I've no idea, if you want to look into it that would be great. Maybe create an issue tracking this as well as it seems like it could be fairly important. |
It turns out that it's because the control flow doesn't handle I've created and internal issue for this |
I'm seeing consistency check failures outside of the tests that are affected by this PR. I think that means something's broken on the particular version of |
I think you're right. I've tried to fix that in #17944. |
I hope it's ok that I merged. The CI is passing now and I'm looking at the |
No problem at all. |
Add more tests for
rust/deadcode
(inspired by real world code), identifying more spurious and missing results.