ConstraintManager source tree fix #137
Conversation
This reverts commit 93b0ead.
| @@ -167,6 +165,17 @@ private AnnotationLocation getCurrentLocation() { | |||
| } | |||
| } | |||
|
|
|||
| // visitorState.getPath().getLeaf() caused NullPointerException on some real project | |||
| // and doesn't always return correct source tree. Using class tree at least shows which | |||
There was a problem hiding this comment.
-
"Using class tree at least shows " is ambiguous -- i.e. the code in this function actually only use class tree when
visitorState.getPath()isnull.
Maybe only add a code comment on the else branch to document using class tree is a workaround for whenvisitorState.getPath()isnull? -
" and doesn't always return correct source tree": do you have minimum code example for that? Should this be an issue instead of a code comment?
| // visitorState.getPath().getLeaf() caused NullPointerException on some real project | ||
| // and doesn't always return correct source tree. Using class tree at least shows which | ||
| // Java source class there is such unsatisfiable error and also doesn't throw exception | ||
| private Tree getCurrentTreeLocation() { |
| } | ||
|
|
||
| private AnnotationLocation getCurrentLocation() { | ||
| private AnnotationLocation getCurrentAnnotationLocation() { |
| if (visitorState.getPath() != null) { | ||
| return visitorState.getPath().getLeaf(); | ||
| } else { | ||
| return visitorState.getClassTree(); |
There was a problem hiding this comment.
Add TODOs for both methods - why hitting this case(else branch)?
There was a problem hiding this comment.
Issue: find a minimum test case
Throw NPE in else branch, and find the project that causes exception.
CharlesZ-Chen
force-pushed
the
master
branch
2 times, most recently
from
2026571
to
d307fc6
Compare
|
Find which projects failed. |
# Conflicts: # src/checkers/inference/model/ConstraintManager.java
|
Sorry I can't recall how I triggered the NPE at that time, but it should be the time when source code is pre-annotated, and the constraints cannot be satisfied before solver (AlwaysFalseConstraint). Maybe you can close this issue for now, and revisit the problem later in an ad-hoc way. I noticed in the latest master branch that source tree location will be passed from |
visitorState.getPath().getLeaf()throwNullPointerExceptionsometimes, and doesn't always return correct source tree location. Change toClassTreeas a temporary fix. There is an issue related to this: typetools#31