about summary refs log tree commit diff
path: root/tvix/eval/src/value/mod.rs
diff options
context:
space:
mode:
authorAdam Joseph <adam@westernsemico.com>2023-12-09T07·25-0800
committerclbot <clbot@tvl.fyi>2023-12-12T14·26+0000
commit8a40f75c2d0cd03e3c3f680f4bd062f0611f2ab8 (patch)
tree871a29a8efbe5c281c412eaded041341bbfba504 /tvix/eval/src/value/mod.rs
parent19d13eb070b99a331477aa84b225c47ea9027211 (diff)
fix(tvix/eval): never use partial_cmp() (partial fix b/338) r/7166
This is part of a fix for b/338.

We should never use PartialOrd::partial_cmp().

All Nix types except floats are obviously totally-ordered.  In
addition, it turns out that because Nix treats division by zero
rather than producing a NaN, and because it does not support
"negative zero", even floats are in fact totally ordered in Nix.

Therefore, every call to PartialOrd::partial_cmp() in tvix is an
error.  We have to *implement* this function, but we should never
call it on built-in types.

Moreover, nix_cmp_ordering() currently returns an Option<Ordering>.
I'm not sure what was going on there, since it's impossible for it
to return None.  This commit fixes it to return simply Ordering
rather than Option<Ordering>.

Change-Id: If5c084164cf19cfb38c5a15554c0422faa5f895d
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10218
Autosubmit: Adam Joseph <adam@westernsemico.com>
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
Diffstat (limited to 'tvix/eval/src/value/mod.rs')
-rw-r--r--tvix/eval/src/value/mod.rs29
1 files changed, 11 insertions, 18 deletions
diff --git a/tvix/eval/src/value/mod.rs b/tvix/eval/src/value/mod.rs
index f4ed7733eec4..c0bdc5fc5eb7 100644
--- a/tvix/eval/src/value/mod.rs
+++ b/tvix/eval/src/value/mod.rs
@@ -600,18 +600,9 @@ impl Value {
     // TODO(amjoseph): de-asyncify this (when called directly by the VM)
     /// Compare `self` against other using (fallible) Nix ordering semantics.
     ///
-    /// Note that as this returns an `Option<Ordering>` it can not directly be
-    /// used as a generator function in the VM. The exact use depends on the
-    /// callsite, as the meaning is interpreted in different ways e.g. based on
-    /// the comparison operator used.
-    ///
     /// The function is intended to be used from within other generator
     /// functions or `gen!` blocks.
-    pub async fn nix_cmp_ordering(
-        self,
-        other: Self,
-        co: GenCo,
-    ) -> Result<Option<Ordering>, ErrorKind> {
+    pub async fn nix_cmp_ordering(self, other: Self, co: GenCo) -> Result<Ordering, ErrorKind> {
         Self::nix_cmp_ordering_(self, other, co).await
     }
 
@@ -619,19 +610,21 @@ impl Value {
         mut myself: Self,
         mut other: Self,
         co: GenCo,
-    ) -> Result<Option<Ordering>, ErrorKind> {
+    ) -> Result<Ordering, ErrorKind> {
         'outer: loop {
             match (myself, other) {
                 // same types
-                (Value::Integer(i1), Value::Integer(i2)) => return Ok(i1.partial_cmp(&i2)),
-                (Value::Float(f1), Value::Float(f2)) => return Ok(f1.partial_cmp(&f2)),
-                (Value::String(s1), Value::String(s2)) => return Ok(s1.partial_cmp(&s2)),
+                // Nix does not support NaN or "negative infinity",
+                // so its floats are in fact totally ordered.
+                (Value::Integer(i1), Value::Integer(i2)) => return Ok(i1.cmp(&i2)),
+                (Value::Float(f1), Value::Float(f2)) => return Ok(f1.total_cmp(&f2)),
+                (Value::String(s1), Value::String(s2)) => return Ok(s1.cmp(&s2)),
                 (Value::List(l1), Value::List(l2)) => {
                     for i in 0.. {
                         if i == l2.len() {
-                            return Ok(Some(Ordering::Greater));
+                            return Ok(Ordering::Greater);
                         } else if i == l1.len() {
-                            return Ok(Some(Ordering::Less));
+                            return Ok(Ordering::Less);
                         } else if !generators::check_equality(
                             &co,
                             l1[i].clone(),
@@ -651,8 +644,8 @@ impl Value {
                 }
 
                 // different types
-                (Value::Integer(i1), Value::Float(f2)) => return Ok((i1 as f64).partial_cmp(&f2)),
-                (Value::Float(f1), Value::Integer(i2)) => return Ok(f1.partial_cmp(&(i2 as f64))),
+                (Value::Integer(i1), Value::Float(f2)) => return Ok((i1 as f64).total_cmp(&f2)),
+                (Value::Float(f1), Value::Integer(i2)) => return Ok(f1.total_cmp(&(i2 as f64))),
 
                 // unsupported types
                 (lhs, rhs) => {