run_parallel_commands doesn't handle exceptions in commands the same way run_commands does

[x4lldux]

When a command crashes during run_parallel_commands, the entire test just crashes also, whilerun_commands catches such a situation and returns a exception tuple as a result. Shouldn't they work the same way?

[kostis]

Please submit a minimal test case that illustrates the issue.

[x4lldux]

I can, but it'll be in Elixir using PropCheck library (a thin wrapper around PropEr); my Erlang is not that strong. But I've traced the origin of it:

• run_parallel_commands uses pmap to run commands in parallel, wrapped in execute,
• execute is using apply BIF instead of safe_apply like run_commands does
• pmap spawns process and awaits the answers,
• spawned process catch any error, while executing the command, and send it as a message to pmap
• when the message, pmap gets, is a crash it just rethrows it

So when a SUT has a crashes, run_commands will return {exception, _, _, _} tuple as a result, while run_parallel_commands just forwards the crash.

[kostis]

I can, but it'll be in Elixir using PropCheck library (a thin wrapper around PropEr); my Erlang is not that strong.

Well, you've came to the PropEr (TM) repository. Here you have your chance to strengthen your Erlang skills ;-)

[x4lldux]

Well, you've came to the PropEr (TM) repository.

Which runs on a BEAM™ that supports many other languages. Here you have your chance to make us feel included ;)

[x4lldux]

@kostis test case added, see commit x4lldux/proper@c932eef62cfddc8a2726d09a2cb9bf5e81fd6015

[x4lldux]

Also, the docs for run_parallel_commands are missing the exception tuple as a valid result type and that's what is being returned when command raises during the sequential part.