Evaluate arbitrary markers to false
#3681
Conversation
crates/uv-resolver/src/marker.rs
Outdated
| @@ -35,7 +35,8 @@ pub(crate) fn is_disjoint(first: &MarkerTree, second: &MarkerTree) -> bool { | |||
| string_is_disjoint(expr1, expr2) | |||
| } | |||
| MarkerExpression::Extra { operator, name } => extra_is_disjoint(operator, name, expr2), | |||
| MarkerExpression::Arbitrary { .. } => false, | |||
| // `Arbitrary` expressions always evaluate to `false`, and are thus always disjoint. | |||
| MarkerExpression::Arbitrary { .. } => true, | |||
There was a problem hiding this comment.
@ibraheemdev -- is this wrong? I was trying to understand the comment here: #3679 (comment)
There was a problem hiding this comment.
Yes, that is correct. However, I wonder if maybe we should be pessimistic regarding forking on meaningless expressions, cc @BurntSushi.
Also if we do make this change, I expected a test to be failing. I think we need to check if expr2 is Arbitrary as well, I can make that change in a later PR.
There was a problem hiding this comment.
Okay, think I fixed this.
There was a problem hiding this comment.
I'm not fully sure. My instinct is also that we probably ought to be pessimistic on forking here. If we say two arbitrary expressions are disjoint, then we will fork. But logically speaking, if the arbitrary expressions are always false, then, well, I guess that makes sense?
I'm happy to let practice dictate what we do here. (So this is fine for now.)
| MarkerExpression::Arbitrary { .. } => true, | ||
| MarkerExpression::Arbitrary { .. } => false, |
There was a problem hiding this comment.
If we always evaluate these to false that means that we will skip any requirement that has unknown marker expressions right? It seems like true would ignore the bad marker which makes more sense?
There was a problem hiding this comment.
"true and warn that it's meaningless" seems like good behavior?
There was a problem hiding this comment.
I believe this was our behavior until very recently (return false here) and is pip's behavior.
There was a problem hiding this comment.
To be explicit: I think we should omit requirements that have invalid markers, not include them.
charliermarsh
force-pushed
the
charlie/arbitrary
branch
from
5db9d95 to
4d08aed
Compare
charliermarsh
force-pushed
the
charlie/arbitrary
branch
from
4d08aed to
18ad21b
Compare
Summary
See: #3679 (comment).
Closes: #3675 (although I think we have another improvement to make there -- will file separately).