Replace url crate with minimal custom parser in gix-url

gix-credentials/src/protocol/context/mod.rs

@@ -101,7 +101,12 @@ mod mutate {
             self.username = url.user().map(ToOwned::to_owned);
             self.password = url.password().map(ToOwned::to_owned);
             self.host = url.host().map(ToOwned::to_owned).map(|mut host| {
-                if let Some(port) = url.port {
+                let port = url.port.filter(|port| {
+                    url.scheme
+                        .default_port()
+                        .is_none_or(|default_port| *port != default_port)
+                });
+                if let Some(port) = port {
                     use std::fmt::Write;
                     write!(host, ":{port}").expect("infallible");
                 }

gix-url/Cargo.toml

@@ -24,7 +24,6 @@ gix-path = { version = "^0.10.21", path = "../gix-path" }
 
 serde = { version = "1.0.114", optional = true, default-features = false, features = ["std", "derive"] }
 thiserror = "2.0.17"
-url = "2.5.2"
 bstr = { version = "1.12.0", default-features = false, features = ["std"] }
 percent-encoding = "2.3.1"
 

gix-url/src/lib.rs

@@ -22,6 +22,9 @@ mod impls;
 ///
 pub mod parse;
 
+/// Minimal URL parser to replace the `url` crate dependency
+mod simple_url;
+
 /// Parse the given `bytes` as a [git url](Url).
 ///
 /// # Note

gix-url/src/parse.rs

@@ -19,7 +19,7 @@ pub enum Error {
     Url {
         url: String,
         kind: UrlKind,
-        source: url::ParseError,
+        source: crate::simple_url::UrlParseError,
     },
 
     #[error("The host portion of the following URL is too long ({} bytes, {len} bytes total): {truncated_url:?}", truncated_url.len())]
@@ -99,30 +99,33 @@ pub(crate) fn url(input: &BStr, protocol_end: usize) -> Result<crate::Url, Error
         });
     }
     let (input, url) = input_to_utf8_and_url(input, UrlKind::Url)?;
-    let scheme = url.scheme().into();
+    let scheme = Scheme::from(url.scheme.as_str());
 
-    if matches!(scheme, Scheme::Git | Scheme::Ssh) && url.path().is_empty() {
+    if matches!(scheme, Scheme::Git | Scheme::Ssh) && url.path.is_empty() {
         return Err(Error::MissingRepositoryPath {
             url: input.into(),
             kind: UrlKind::Url,
         });
     }
 
-    if url.cannot_be_a_base() {
-        return Err(Error::RelativeUrl { url: input.to_owned() });
-    }
+    // Normalize empty path to "/" for http/https URLs only
+    let path = if url.path.is_empty() && matches!(scheme, Scheme::Http | Scheme::Https) {
+        "/".into()
+    } else {
+        url.path.into()
+    };
 
     Ok(crate::Url {
         serialize_alternative_form: false,
         scheme,
         user: url_user(&url, UrlKind::Url)?,
         password: url
-            .password()
+            .password
             .map(|s| percent_decoded_utf8(s, UrlKind::Url))
             .transpose()?,
-        host: url.host_str().map(Into::into),
-        port: url.port(),
-        path: url.path().into(),
+        host: url.host,
+        port: url.port,
+        path,
     })
 }
 
@@ -156,31 +159,32 @@ pub(crate) fn scp(input: &BStr, colon: usize) -> Result<crate::Url, Error> {
     // should never differ in any other way (ssh URLs should not contain a query or fragment part).
     // To avoid the various off-by-one errors caused by the `/` characters, we keep using the path
     // determined above and can therefore skip parsing it here as well.
-    let url = url::Url::parse(&format!("ssh://{host}")).map_err(|source| Error::Url {
+    let url_string = format!("ssh://{host}");
+    let url = crate::simple_url::ParsedUrl::parse(&url_string).map_err(|source| Error::Url {
         url: input.to_owned(),
         kind: UrlKind::Scp,
         source,
     })?;
 
     Ok(crate::Url {
         serialize_alternative_form: true,
-        scheme: url.scheme().into(),
+        scheme: Scheme::from(url.scheme.as_str()),
         user: url_user(&url, UrlKind::Scp)?,
         password: url
-            .password()
+            .password
             .map(|s| percent_decoded_utf8(s, UrlKind::Scp))
             .transpose()?,
-        host: url.host_str().map(Into::into),
-        port: url.port(),
+        host: url.host,
+        port: url.port,
         path: path.into(),
     })
 }
 
-fn url_user(url: &url::Url, kind: UrlKind) -> Result<Option<String>, Error> {
-    if url.username().is_empty() && url.password().is_none() {
+fn url_user(url: &crate::simple_url::ParsedUrl<'_>, kind: UrlKind) -> Result<Option<String>, Error> {
+    if url.username.is_empty() && url.password.is_none() {
         Ok(None)
     } else {
-        Ok(Some(percent_decoded_utf8(url.username(), kind)?))
+        Ok(Some(percent_decoded_utf8(url.username, kind)?))
     }
 }
 
@@ -269,13 +273,22 @@ fn input_to_utf8(input: &BStr, kind: UrlKind) -> Result<&str, Error> {
     })
 }
 
-fn input_to_utf8_and_url(input: &BStr, kind: UrlKind) -> Result<(&str, url::Url), Error> {
+fn input_to_utf8_and_url(input: &BStr, kind: UrlKind) -> Result<(&str, crate::simple_url::ParsedUrl<'_>), Error> {
     let input = input_to_utf8(input, kind)?;
-    url::Url::parse(input)
+    crate::simple_url::ParsedUrl::parse(input)
         .map(|url| (input, url))
-        .map_err(|source| Error::Url {
-            url: input.to_owned(),
-            kind,
-            source,
+        .map_err(|source| {
+            // If the parser rejected it as RelativeUrlWithoutBase, map to Error::RelativeUrl
+            // to match the expected error type for malformed URLs like "invalid:://"
+            match source {
+                crate::simple_url::UrlParseError::RelativeUrlWithoutBase => {
+                    Error::RelativeUrl { url: input.to_owned() }
+                }
+                _ => Error::Url {
+                    url: input.to_owned(),
+                    kind,
+                    source,
+                },
+            }
         })
 }

gix-url/src/scheme.rs

@@ -48,6 +48,17 @@ impl Scheme {
             Ext(name) => name.as_str(),
         }
     }
+
+    /// Return the default port for this scheme, or `None` if it is not known.
+    pub fn default_port(&self) -> Option<u16> {
+        match self {
+            Scheme::Http => Some(80),
+            Scheme::Https => Some(443),
+            Scheme::Ssh => Some(22),
+            Scheme::Git => Some(9418),
+            _ => None,
+        }
+    }
 }
 
 impl std::fmt::Display for Scheme {