Hi Rob, Thanks very much for your help. I’ve found some errors in my build process, which casts doubt over the how that assert statement was reached. I’m still getting erroneous behaviour from my code when I exceed 2GB read / write buffers per process, so there may be something more to this. I’ll continue working and get back to you if I can isolate a useful test case. Thanks again, Sean From: Latham, Robert J. [mailto:[email protected]] Sent: 29 October 2014 14:58 To: Sean Delaney; discuss Subject: Re: discuss Digest, Vol 24, Issue 35 I probably just need to ensure lustre picked up all changes from common code. Can you confirm your code works if you prefix the filename with "ufs:" ? ==rob -- sent from a phone From: Sean Delaney <[email protected]<mailto:[email protected]>> Sent: Oct 29, 2014 9:54 AM To: [email protected]<mailto:[email protected]> Cc: Latham, Robert J. Subject: RE: discuss Digest, Vol 24, Issue 35 Hi Rob, Yes, this is mpich-3.1.3, as downloaded from the website in the last 2 days. I was alerted to the buf_idx[] overflow by a failing assert statement, which checks for overflow when assigning values to buf_idx[]. I'm running on lustre. I can't provide my current code as a test case. I may be able to prepare something later, but I need to focus on making my code work first. These variables have type 'int', whereas something like 'size_t' or 'ADIO_Offset' might be more appropriate to store buffer offsets, where constraints are not known. Sean ----------------------- this is in the latest 3.1.3 MPICH? I thought I bumped up the size of buf_idx and related fields, though my test cases are meager at best. Do you have a test case? Which file system? ==rob ______________________________________________________________________ This email has been scanned by the Symantec Email Security.cloud service. For more information please visit http://www.symanteccloud.com ______________________________________________________________________ ______________________________________________________________________ This email has been scanned by the Symantec Email Security.cloud service. For more information please visit http://www.symanteccloud.com ______________________________________________________________________ ______________________________________________________________________ This email has been scanned by the Symantec Email Security.cloud service. For more information please visit http://www.symanteccloud.com ______________________________________________________________________ _______________________________________________ discuss mailing list [email protected] To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss