Test run black on self #3114
The only main advantage I think we run these as a unites is that when they fail, their stack trace might make it easier identify the bug and fix rather than the output that pre-commit running black itself generates.
There are benefits to running each way. Minimal, but there are benefits, as mention above.
Let me ask you: What is the benefit you see of running black on itself via
Hey @cooperlees ,
The reason I don't think that running
This is something that
I think that the main advantage of running black via the command line on itself is that it's our first opportunity to see how Black works end-to-end. You get few things out of that:
I don't think that we get any added value by running Black as a unit test on each source file. I think that if we run Black end-to-end on itself we can get much more information.
I hope this convinces you. If not, I'll close this PR and create a new one with the change you requested.
I'm sorry if I came a bit too strong, I really want the best for Black and I'm here to help. This project is really important to me.
This does seem the case after reading docs + testing locally. Due to that I feel we remove
I personally now don't see the need to run the latest released black in our CI runs. I feel we only want to catch developers breaking
I'm well aware of the value of this kind of test here (I do this for many other projects I maintain as an integration / system test - e.g. bandersnatch). There is no need to explain that. I just thought we already had this and you were duplicating this. It does seems