mkraska
commented
4 years ago I want to keep a minimal footprint in terms of storage space and bandwidth. Therefore I limit the repository to the source files and those result files (mainly images) required for documentation.
Yet I understand that some users might not get CGX to work, but still want to test CCX. I could add the required files for some selected examples. Which candidates would you recommend for this? Would the complete CCX input be sufficient or would you also want the FRD files?
As to git committing: I am not sure that I get your point right:
I never delete any file just for committing something new. Instead, I add the new or updated files explicitly using git add. I am not a git power user and there might be more efficient approaches. I guess things could get a lot easier with proper use of .gitignore.
In the context of automatic regression testing I thought of adding some reference results to the repository for automatic diff check with new CGX/CCX versions or as local sanity check for new users. This is what the test.py scripts finally should do. Not just run the simulation but also check for differences to the reference results.
imirzov
Dear Prof. Kraska,
Each time a user runs test.py he produces many output files. Why don't you keep all output files on the GutHub repository?
I suppose it could be useful: for example, if someone fail to successfully run an example and generate all the model files, he still can use models directly from the repository.
Also keeping all the files on the GitHub is useful for developers. Now before committing a new feature a have to delete all generated output files. It's bothering.