[Enhanced Switch][Null] Missing Null pointer access warning with total/unconditional patterns in case labels #3382
Conversation
srikanth-sankaran
commented
Dec 3, 2024
srikanth-sankaran
commented
- Fixes [Enhanced Switch][Null] Missing Null pointer access warning with total/unconditional patterns #3381
|
Expecting some tests to fail org.eclipse.jdt.core.tests.compiler.regression.NullAnnotationTests21.test_totalTypePatternDoesNotAdmitNull() but these need to be studied to see if expectation is valid |
|
Looking at So when @stephan-herrmann - The proposed patch is still WIP, comments are welcome |
srikanth-sankaran
force-pushed
the
Issue3381
branch
from
9821164 to
2ee0b1e
Compare
|
I think I have remastered the "failing" tests with right expectations. @stephan-herrmann thanks for your review. |
total/unconditional patterns in case labels * Fixes eclipse-jdt#3381
srikanth-sankaran
force-pushed
the
Issue3381
branch
from
2ee0b1e to
a753a21
Compare
looks like this changed between Java 18 and 19.
looking into it now. |
| ---------- | ||
| 1. ERROR in X.java (at line 6) | ||
| switch (i) { | ||
| ^ | ||
| Null pointer access: The variable i can only be null at this location | ||
| ---------- | ||
| 2. ERROR in X.java (at line 12) | ||
| switch (i) { | ||
| ^ | ||
| Null pointer access: This expression of type Integer is null but requires auto-unboxing | ||
| ---------- |
There was a problem hiding this comment.
@srikanth-sankaran do you have an idea why these are different error messages?
There was a problem hiding this comment.
@srikanth-sankaran do you have an idea why these are different error messages?
Yes, I had looked into this. In the former case its is a type switch which doesn't call for unboxing and in the latter switch on the unboxed integer. This results in a special message in the latter case.
total/unconditional patterns in case labels + code simplification Fixes eclipse-jdt#3381
|
@srikanth-sankaran your turn again :) |
|
Your additional changes look good, Thanks @stephan-herrmann! |
eclipse-jdt#3319 This was fixed by eclipse-jdt#3382
|
The first commit from here also fixes #3319 - basically the isNullHostile method's wrong result in the presence of total patterns was inhibiting a diagnostic from being reported. |
