Neil Fortner

Forgot to add: If all ranks read the entire file then the program passes. This implies that the complicated datatype on rank 1 is interfering with I/O on other ranks...

> mpirun --mca io ^ompio -n 3 a.out Test passes with ROMIO, looks like the problem is with OMPIO. Thanks!

We've confirmed the test program works on the latest snapshot. Thanks! We'll test the HDF5 branch in question with it soon to verify.

The multi dataset branch of HDF5 that was failing with 4.1.2 passes with the latest master. Thanks again!

We've noticed a bug in mpich 4.0-4.0.3 using the default "ch4:ofi" device. Last we checked it had been fixed in the 4.1 branch. If you need to use 4.0-4.0.3 you...

This is a problem with the compiler, but we'll leave this issue open until the compiler is fixed.

You shouldn't see failures with threadsafe enabled, unless one of the threads is changing assumptions made by the other threads (such as resizing the dataset). Do you have an example...

What errors do you see?

And are you only extending the datasets or also shrinking them (along any dimension)?

For now, passthrough VOL connectors should avoid doing anything with the file in the open and create callbacks except opening it. If connectors need to do anything else they should...