16639, “bradcray”, “Scripting to help diagnose IB / portability problems?”, “2020-10-29T22:50:36Z”
This is a capture of an idea I’ve been wondering about recently: Given the challenges that so many IB / cluster users seem to have getting up and running on their system, I wonder whether we could create something of a gentle ramp of checks that would make sure things are working as expected and provide us / them with more detail when something doesn’t. For example, imagine a script you could run which would try the following in order:
- compiling and launching a single-node C “Hello world”
- compiling and launching a single-node GASNet “Hello world”
- compiling and launching a multi-node GASNet “Hello world”
- compiling and launching a single-node Chapel “Hello world”
- compiling and launching a multi-node Chapel “Hello world”
The idea is that each of these would be compiled using the same C compiler and launcher as Chapel would, and that if any of the steps failed, it would print out key information like which C compiler, which launcher, the --verbose output, printchplenv output, and such. This might help us (or help users figure out) where things went off the rails (where today, we effectively jump directly to step 5).