question about running llvm test-suite on remote machine


I’ve looked into use of the remote variables: REMOTE_HOST, REMOTE_CLIENT, REMOTE_USER, etc. Can you tell me how exactly one uses these variables to run the llvm test-suite remotely?

As far as I can tell, the local host and remote host directory structures have to match exactly. In addition, it seems that the remote host has to pre-generate building the test executables, while the local host just executes them. Is that correct? If that is the case, I’m not sure what the benefit is.

I was expecting that the local host running the tests would compile the binaries and then somehow transfer them to the remote host for execution which is something that dejagnu seems capable of doing. It seems that the llvm test-suite is doing something different. Perhaps you are using shared network drives between local and remote hosts which would make this type of setup work. It also seems like this is not documented on how to do remote testing.

If someone could tell me how this should be used, I’d greatly appreciate it.



Jose Rangel

QA Engineer

Arxan Technologies, Inc.

Your guesses are correct. All the files must reside on a file system shared between the two machines, and they must be mounted so that the same paths work on both.

The tests are compiled on the local machine, and the various REMOTE settings are then used to run the resulting binaries on the remote machine. For example, if you set REMOTE_CLIENT=ssh, the tests will be run via ssh on REMOTE_HOST.

It should not be too hard to transfer the files to the remote host and then transfer the output back, but no one has done that yet. If you’re interested in working on that, it would be much appreciated.

Hi Bob,

Thanks a lot for the explanation. I think I’ll try using the shared directory for now, but I’ll keep the other capability in mind since it would be cool to have.