search

ReactionMechanismGenerator / RMG-tests

Continous Integration Testing Platform for RMG-Py
3 stars 14 forks source link

Switch to GitHub actions #79

Closed sevyharris closed 2 years ago

sevyharris commented 2 years ago

Move regression and integration tests from Travis CI to Github Actions

rwest commented 2 years ago

A general comment: it would be great if you could briefly describe why you're doing something in your commit messages.

An ideal commit message (IMHO) is something like

One line (<80 char) title of what you did

After a blank line, then have a few lines (with manually inserted line breaks at about 80 characters) describing WHY you did the thing. You could add a little more detail about WHAT you did, but one could argue that too much detail is redundant, as the diff shows precisely what you did. WHY you did it, cannot in general be guessed.
The more obvious the reason, the shorter the explanation required. But you never know who's going to be reading this later and how much context they have.

sevyharris commented 2 years ago

Did the ln -sfn $(which python-jl) $(which python) part of the julia installing not work? I'm guessing that what that line was trying to do is link python to python-jl. Your method of explicitly calling python-jl might be preferable, but I'm curious why the link method didn't work (and if it should be removed if it can't be made to work)

I don't think it was the linking that failed. Changing the python calls to python-jl was just one of several attempts to get the Julia packages working. At this point, we think it was a version compatibility issue with StochasticDiffEq, so hopefully downgrading that version will fix rms. I'll ask whether we should keep the python-jl calls in.

rwest commented 2 years ago

This seems quite tricky. Has @mjohnson541 been helping with the Julia and RMS stuff?

sevyharris commented 2 years ago

This seems quite tricky. Has @mjohnson541 been helping with the Julia and RMS stuff?

Yes, he's the one who suggested downgrading the StochasticDiffEq version. I filed an issue here to ask for some extra help. What's frustrating is that I can get it installed and running on toastie by trying the install again after it fails. I might be able to replicate the process in Github Actions, but I'm probably doing something wrong if it's failing in the first place.

sevyharris commented 2 years ago

Rebased commit history to be much neater. I'm testing this now, so if this run fails after ~90m, it should be ready to merge.