pathauto handles language-neutral aliases wrong resulting in duplicates

I confirm that this patch works as described. Thanks to this change I no longer end-up with colliding aliases when a language-neutral alias is being created while an identical language-specific one already exists. path_pathauto_is_alias_reserved() now properly flags the proposed alias as being reserved which leads to pathauto_alias_uniquify() testing iterations until a unique suffix is found to prevent collisions, as expected.

Thanks for the fix!

The only change this makes vs the original is that it comment has been reformatted to 80 chars, per the Drupal coding standards.

I am not sure why more people are not reporting this problem. Patch #5 fixes it for me.

+++ b/pathauto.module
@@ -454,12 +454,17 @@ function pathauto_is_alias_reserved($alias, $source, $langcode = LANGUAGE_NONE)
+  $sql = "SELECT pid FROM {url_alias} WHERE source <> :source AND alias = :alias";
+  $tokens = array(':source' => $source, ':alias' => $alias);
+  if ($langcode != LANGUAGE_NONE) {
+    $sql .= " AND language IN (:language, :language_none)";
+    $tokens += array(':language' => $langcode, ':language_none' => LANGUAGE_NONE);
+  }  ¶
+
+  return (bool) db_query_range($sql, 0, 1, $tokens)->fetchField();

This patch shoud definitely be converting this to db_select() now. I don't want to be have SQL strings constructed together with strings like this.

Patch rewritten to follow db_select() standards.

Fixes use of $langcode, rather than $language.

Hi,

I have the same pb with the recommended release 7-1.2

I try the 7-1.x-dev and same bug, and when I apply successfully your patch, it's work

Thank a lot,

I hope the dev version is stable because I put it in production.

Sim

Thanks for the review @kiricou. Can anyone else review so we can RTBC this?

This needs a test case to confirm it works and will not regress in the future.

Setting to needs work per #21

Here's a revised patch that adds a test to the end of the testLanguageAliases() method. It ensures that when a language neutral node is saved with what will end up being the same alias as a node with its language defined, the alias suffix is incremented correctly.

Indentation is off lines 728 & 729 :)

Otherwise patch looks good.

Ah, you're right! Here are revised patches with the indentation fix.

I noticed my comment was wrong for the test. the -1 is a suffix, not a prefix.

Thanks, patch looks good now. If anyone else can review this can be RTBC'ed

Complete patch in #30 works for me. RTBC'ing based on this and #34.

Patch #30 (complete version) also works for me. Thanks.

Works for me too, thanks.

Committed #30 to 7.x-1.x. Thanks!