Bugzilla – Full Text Bug Listing
| Summary: | VerifyError due to bogus lowering in inferred callout-to-field | ||
|---|---|---|---|
| Product: | [Tools] Objectteams | Reporter: | Stephan Herrmann <stephan.herrmann> |
| Component: | OTJ | Assignee: | Stephan Herrmann <stephan.herrmann> |
| Status: | VERIFIED FIXED | QA Contact: | |
| Severity: | normal | ||
| Priority: | P3 | ||
| Version: | 2.0 | ||
| Target Milestone: | 2.2 M2 | ||
| Hardware: | Other | ||
| OS: | Linux | ||
| Whiteboard: | |||
VerifyError is bad, schedule for M2. This was closed by accident. The error can still be reproduced. This is what caused the error: - lowering of the BlockScope arg caused Config.requireTypeAdjustment() which triggered the InsertTypeAdjustmentsVisitor - inside InsertTypeAdjustmentsVisitor a lowering conversion was inserted for the fieldReference this.expression The fix consists of one additional check: - don't add lowering conversion for a field reference that already has a syntheticAccessor w/ purpose InferredCalloutToField Released for 2.2 M2 via commit 47700cc09e0d3152088b67bbb6920af31adc19d2. Verified using 2.2 M5 |
In my null annotation implementation I had code like this protected class ReturnStatement extends NodeWithBits playedBy ReturnStatement { Expression getExpression() -> get Expression expression; FlowInfo analyseCode(BlockScope currentScope, FlowContext flowContext, FlowInfo flowInfo) <- replace FlowInfo analyseCode(BlockScope currentScope, FlowContext flowContext, FlowInfo flowInfo); callin FlowInfo analyseCode(BlockScope currentScope, FlowContext flowContext, FlowInfo flowInfo) { if (this.expression != null) { this.expression.analyseCode(currentScope, flowContext, flowInfo); ... The generated code triggered a VerifyError because the inferred c-t-f for this.exression (last line) bogusly tried to lower the receiver (this) to the base version of ReturnStatement. It seems the compiler couldn't decide whether to use the existing explicit c-t-f (use this) or to directly access the base field (use lower(this)). OTOH, also lowering for the argument currentScope had a finger in the pie (workaround: make this lowering explicit). This could indicate that the loweringRequired flag was consumed at a wrong location. Might/might not be related to bug 352914 or bug 353894 comment 13.