You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently pointerescapes tracks a single boolean value across the callgraph. It marks each pointer argument it finds as 'escapes' or not.
There are several other properties of these pointers we want to track to handle other bad pointer usage like pointer comparisons. To be able to do so without a lot of duplication, we need to extend the boolean value being tracked to an n-tuple of boolean values. The API design should be extensible enough to allow us to add extra flags to track as needed.
Once this is in place, we can tie that n-tuple into analyzers that are focused on the function declarations only, and trust the callgraph analyzer to perform whatever induction is needed for us.
Once this is done, pointerescapes will be a misnomer, so maybe it should be refactored into an induction analyzer under the callgraph package.
The text was updated successfully, but these errors were encountered:
kalexmills
changed the title
Update pointerescapes to work over a product domain
Update the induction in pointerescapes to work over an extensible product domain
Jan 9, 2021
Currently
pointerescapes
tracks a single boolean value across the callgraph. It marks each pointer argument it finds as 'escapes' or not.There are several other properties of these pointers we want to track to handle other bad pointer usage like pointer comparisons. To be able to do so without a lot of duplication, we need to extend the boolean value being tracked to an n-tuple of boolean values. The API design should be extensible enough to allow us to add extra flags to track as needed.
Once this is in place, we can tie that n-tuple into analyzers that are focused on the function declarations only, and trust the callgraph analyzer to perform whatever induction is needed for us.
Once this is done,
pointerescapes
will be a misnomer, so maybe it should be refactored into aninduction
analyzer under thecallgraph
package.The text was updated successfully, but these errors were encountered: