You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
At a minimum we should verify that the things we output compile. This sounds doable-if-tedious. We do this now for some things, in the benchmarks. (It probably should be wider, although I'm not going to say we should try every flag combo or anything -- just ones likely to interact).
Ideally we could go one step further and verify that the data we output matches some reference version of the data. ucd-generate itself could even emit the tests.
The text was updated successfully, but these errors were encountered:
In theory I'm okay with this. But I personally don't want to have too much ceremony around it. That is, I would like to keep the barrier to adding new sub-commands relatively small. So tests that don't need to be changed or added to when a new command is added would be a nice win. (e.g., Checking that the output compiles.)
At a minimum we should verify that the things we output compile. This sounds doable-if-tedious. We do this now for some things, in the benchmarks. (It probably should be wider, although I'm not going to say we should try every flag combo or anything -- just ones likely to interact).
Ideally we could go one step further and verify that the data we output matches some reference version of the data. ucd-generate itself could even emit the tests.
The text was updated successfully, but these errors were encountered: