Fix#6784 - Add diagnostic for for-loop update expressions with no effect #6832
Conversation
There was a problem hiding this comment.
The name based checking is a big code smell, and always should be avoided.
The right way to check is to implement a general side effect detection logic in post lowering validation phase next to where we are detecting missing returns. If we see a loop whose condition is a constant and the loop does not have any exit jumps, then there should be a warning.
@csyonghe by post lowering validation does that mean |
|
Correct, we should add another pass after the check missing return pass to validate this. |
|
Okay. Would require to redo the logic. |
This commit introduces a new warning diagnostic (30506) for cases where the update expression in a for-loop does not modify any variables, potentially leading to infinite loops. The implementation includes checks in the `SemanticsStmtVisitor` to identify such expressions and provide appropriate warnings. Additionally, tests have been added to verify the new diagnostic behavior.
This reverts commit d709129.
3d9367c to
f400187
Compare
8bffd28 to
3e2bee9
Compare
| @@ -56,5 +58,9 @@ float doSomething(int x) | |||
| { | |||
| i--; | |||
| } | |||
| for (int i = 1; i < 10; i+1) // warn. | |||
There was a problem hiding this comment.
@csyonghe - I observed that Diagnostics::potentialInfiniteLoop is only emitted for first occurence in doSomething. In the code we are doing a graph search - so ideally this should be caught, but i am not able to wrap my head on this.
| void checkForPotentialInfiniteLoops(IRModule* module, DiagnosticSink* sink, bool diagnoseWarning) | ||
| { | ||
| // Look for loops in the module | ||
| checkForPotentialInfiniteLoopsInInst(module->getModuleInst(), sink, diagnoseWarning); |
There was a problem hiding this comment.
Should this be done only for any IRLoop in the module, instead of just starting off on the first inst?
There was a problem hiding this comment.
We start procesing all IR instructions - The IRloop check is done inside this function at L105
|
The original issue is NOT giving warning -> However able to do on sample repro loop_issue1.slang Image-2 |
This commit introduces a new warning diagnostic (30506) for cases where the update expression in a for-loop does not modify any variables, potentially leading to infinite loops. The implementation includes checks in the
SemanticsStmtVisitorto identify such expressions and provide appropriate warnings. Additionally, tests have been added to verify the new diagnostic behavior.