JS: model property projection calls #113
Conversation
ghost
self-requested a review
as a 
| ( | ||
| callee = DataFlow::moduleMember("dotty", "search") and | ||
| objectIndex = 0 and | ||
| selectorIndex = 1 |
| import javascript | ||
|
|
||
| /** | ||
| * A call that projects selected properties of an object. |
There was a problem hiding this comment.
Here and below: these comments could do with examples of what is being modelled.
| abstract DataFlow::Node getASelector(); | ||
|
|
||
| /** | ||
| * Holds if this call returns the value of a single projected property. |
There was a problem hiding this comment.
This is particularly arcane; please explain more carefully and add an example.
|
Comments addressed. |
|
Squashed the fixups. |
|
The tests actually seem to have succeeded, but the status didn't propagate (@pavel). Merging. |
|
@esben-semmle, this PR seems to have been left behind when creating https://github.com/Semmle/ql/tree/rc/1.18. Could you open a PR to cherry-pick the commits across? Same for #115. |
ghost
mentioned this pull request
|
@xiemaisi Thanks for notifying me, but it does not seem that I'm involved in this project in any way. https://blog.github.com/2011-03-23-mention-somebody-they-re-notified/ |
|
Oops, sorry, wrong Pavel... (It's a great project, though! 😄) |
Kotlin: Allow building only a single way
fix implicit this
…this fix implicit this
This PR adds a taint step for property reads done through calls that projects selected properties of an object.
The property projections that have been modeled have slightly different features. So I have adopted @asger-semmle's extensible architecture proposal immediately. We may want to discuss how we generally document the intended usage of the architecture, I have simply added an explanation to the module docstring for now.
This PR only makes use of the object-argument in the models.
For the record, the libraries support various features for the selector-argument:
'a.b[42].c'['a', 'b', 42, 'c']'a.\*.c'We can consider implementing support for these once our general heap treatment is better.
An evaluation with the security suite on big-apps.slugs flags nothing new, and it does not have a significant performance impact (internal link).