tnid does not updated to zero after node delete

+++ b/modules/translation/translation.moduleundefined
@@ -420,7 +420,7 @@ function translation_remove_from_set($node) {
-    if (db_query('SELECT COUNT(*) FROM {node} WHERE tnid = :tnid', array(':tnid' => $node->tnid))->fetchField() == 1) {
+    if (db_query('SELECT COUNT(*) FROM {node} WHERE tnid = :tnid AND nid <> :nid', array(':tnid' => $node->tnid, ':nid' => $node->nid))->fetchField() == 1) {
       // There is only one node left in the set: remove the set altogether.

The comment under the query indicates the intention of the query was to include the node itself.

Also most probably applies to Drupal 8, should be fixed there first, then backported to Drupal 7 (and possibly to Drupal 6 if applies there).

Sending for testing anyway. How come existing tests do not cover this?

Looks generally like a good direction. I think its fine to keep it as its own test.

+++ b/core/modules/translation/lib/Drupal/translation/Tests/TranslationTest.phpundefined
@@ -242,6 +242,58 @@ class TranslationTest extends WebTestBase {
+   *
+   * @todo Improve and enclose in t() the messages.

Test assertion messages should not be t()-ed anymore in Drupal. You should skip doing this.

+++ b/core/modules/translation/lib/Drupal/translation/Tests/TranslationTest.phpundefined
@@ -242,6 +242,58 @@ class TranslationTest extends WebTestBase {
+    // The translation source is not the english node,
+    // this require a proper tnid handling in createTranslation().

I don't understand this comment at all. Also, it does not wrap properly.

+++ b/core/modules/translation/lib/Drupal/translation/Tests/TranslationTest.phpundefined
@@ -242,6 +242,58 @@ class TranslationTest extends WebTestBase {
+    // In this case maybe random which node was selected to translation source.
+    // @todo Write asserts to cover.

What is random?

Also, complete that todo :)

Still interested in working on this?

Related: #1645156: URL generation only works on port 80 when using domain based language negotation (that is about generation not detection). (Posted on the wrong issue).

Fixed the comments as per #7. Appears that the @todo was already done.

Removed an added t() function.

I have reviewed this patch. Lets wait for the bots tests.

okay - that last minor change was a little too eager. Here goes, with quotes.

The latest patch looks good. Fix makes sense and it is thoroughly tested.

Adding sprint tag.

Committed and pushed to 8.x. Thanks!

Did not have the backport tag, but need to be backported (http://drupal.org/node/1688144#comment-62426529).

Looking at #1305378-34: Tokens should use $options['langcode'] and not need a language object and #322995-62: Provide a distinct administration user interface language option test results, maybe some problem has been introduced with this commit?

A proof.

Ok, confirmed this was breaking testbot, so I've rolled this patch back.

The problem (at least on localhost) was that it was still failing the final "Translation set is terminated." test introduced by this patch. Not sure if one of the other ones I committed today conflicted or what.

Try this.

In other words (after discussion with @chx), this did not work anymore due to node_load() changes in http://drupal.org/node/1744046 (depreciated $conditions argument removed), which was committed before this one.

Brace yourself for more of such conflits as APIs change on an ever higher pace and developers find it harder and harder to keep up :/

Did not apply, rerolled.

Back to RTBC then. Thanks @chx for your help!

What about the existing data suffering from this bug? It should be able to provide an upgrade path to find the nodes that are in a translation set all of their own and mark those back to 0 no?

Yes, that makes sense.

I don't think this makes sense to keep on the sprint, given that we target the removal of this making this patch be stripped out of Drupal 8 anyway. It was targeted at Drupal 8 purely as an application of process. If we need someone to spend more time on this (and people don't seem to be flocking here to do it), then it should not take away resources from the people who are trying to focus on meaningful D8MI contributions, which is what the D8MI and sprint tag combinations are for. Honestly at this point it is not easy to tell if this node-copy translation functionality will be removed for sure (still 87 days to go until feature freeze), so not turning this back to Drupal 7, but if that happens, this should be turned back there as being not applicable to Drupal 8. If that does not happen, we need to keep it on Drupal 8 and fix the bug there.

Removing D8MI tag as well in anticipation of the underlying functionality being removed from core.

I'm working on the upgrade path.

Cross-post.

D7 patch worked for me, but I was wondering why it's still not included in Drupal 7?

Did some checking and this is no longer relevant for Drupal 9 but still is for Drupal 7. Moving back to the D7 issue queue.