Make the override warnings prettier.

sdk/lib/_internal/pub/lib/src/source/path.dart

Index: sdk/lib/_internal/pub/lib/src/source/path.dart
diff --git a/sdk/lib/_internal/pub/lib/src/source/path.dart b/sdk/lib/_internal/pub/lib/src/source/path.dart
index af752d674308146e3c0417f5957fafb1b77940b4..76892e1e6528a26573858f4073dff48b9382476a 100644
--- a/sdk/lib/_internal/pub/lib/src/source/path.dart
+++ b/sdk/lib/_internal/pub/lib/src/source/path.dart
@@ -113,6 +113,16 @@ class PathSource extends Source {
     return description;
   }
 
+  /// Converts a parsed relative path to its original relative form.
+  String formatDescription(String containingPath, description) {
+    var sourcePath = description["path"];
+    if (description["relative"]) {
+      sourcePath = path.relative(description['path'], from: containingPath);

[nweiz, 2013/11/26 21:14:46]

I think this should be relative to the pwd, not [containingPath]. Most of the
time the two will be the same, but if pub is run from a subdirectory of the
package I think it'll be clearer if the relative path is pointing to a path
that's resolvable from their current location. Similarly, I think it should
relativize the path even if it's absolute in the pubspec.

The purpose of this output isn't to convey what's in the pubspec, it's to convey
what the dependency is.

[Bob Nystrom, 2013/11/26 22:29:38]

I tried that, but it feels and looks wrong to me. I understand what you're
getting at, but I think it's clearest to show the path that the user specified
in their pubspec. I think they'll understand it's relative to the pubspec
because that's what it was when they authored it.

Showing a different calculated path is confusing because they may wonder where
that strange relative path came from, and it isn't obvious that it's just the
output trying to be friendly.

[nweiz, 2013/11/27 01:11:39]

How about this: if [containingPath] isn't [path.current], we add "(from
${path.relative(containingPath)})" to the end of [sourcePath]?

[Bob Nystrom, 2013/12/03 19:10:55]

The string returned by this is already used in a large string that mentions the
source name, so that would result in something like:

- foo (from path ../blah (from blah))

I really think just showing the pubspec relative path is the cleanest thing to
do here. Given that we can't even run this code without the cwd being the same
as the pubspec directory, I don't think it's worth doing anything different
here.

+    }
+
+    return sourcePath;
+  }
+
   /// Ensures that [description] is a valid path description and returns a
   /// normalized path to the package.
   ///