I'm finding a code which makes the case Te_TyArray in ML_(sizeOfType) to blowup in an assert. I'm attaching a suggested patch which would soften things out and allow the execution to continue. Anyway, the problem is likely related with the debug info reader, since I'm getting the famous "chase_cuOff: no entry for ..." I'm using a pretty complex code, but if you're interested I can attach it as well or try to find a minimal code causing the problem. Best, Antonio J. Peña Barcelona Supercomputing Center Reproducible: Always Steps to Reproduce: 1. Compile MiniMD from mantevo.org 2. Execute valgrind with --read-var-info=yes Actual Results: Assert failed. Execution aborted. Expected Results: Try to continue execution.
Created attachment 95443 [details] Suggested workaround
BTW, this is current trunk, 3.12.0.SVN (r15725). --Antonio
Please could you provide a reproducible test case? The proposed patch itself looks plausible.