Is there any reason folks can think of for why we ban

--changed-since

and normal CLI args? What if you want to say

all changed files and also these 3 tests I want to run

When I first read this, I was like "Oh, why would someone ever do that - that feels weird" But then, I grokked it as

changed-since

is just a shortcut for a list of files to save typing, and then on top I'm just adding a couple more files to the end of a list.... So this made me think "yeah, of course that should work - it would be weird if it didn't" Emotional rollercoaster...

I think the idea of us eagerly erroring right now is we're trying to help the user? I suspect a common issue would be

./pants --changed-since=HEAD test ::

, which means the

--changed-since

is useless... But maybe we fix that via clearer docs, and hoping the user can intuit "wait what, why is it running on so much?" Wdyt?

I've accidentally done

./pants --changed-since=HEAD test ::

before, and the error was helpful - and it took half a second to realize my

::

habit was in overdrive. However, I think the intuitiveness is more around:

./pants --changed-since=HEAD a.py b.py test

vs

./pants test --changed-since=HEAD a.py b.py

imo, it would be good to hold onto this restriction until other refactorings of target roots are completed… it would be good to avoid complicating the interplay

Okay sounds good. The restriction doesn't block me, only realized how easy it would be to lift it

I think we can even wait until there's a request or two. Until someone needs us to change the behavior or report it's too confusing as-is, no need to introduce volatility for everyone in service of a hypothetical.

People have occasionally asked for this behavior, and it does make sense to have I think

Yea... I think that someone asked recently whether it should compute the intersection