Joe Thornber

Update submitted atomic in submit_all()

1

If the user calls submit_all() directly, the submitted atomic is not updated. Leading to a crash. I suspect most users rely on waiting on a completion to implicitly submit so...

We get a duplicate write lock, and the metadata is now corrupt (even though we hadn't committed). cache_restore is the only cache tool that writes data, and it always starts...

If do 'vaf' to select the current function, and then :ChatGPTRun docstring, then I get just a comment rather than a comment + the original function/context.