class isn't set in FETCH_OBJECT when class_name isn't set

I run tests

Can someone add testing for this or can post a query where this occurs.

Hi, ran into the same problem when using ultimate_cron too, agree the function signature implies you can pass the class_name for StatementPrefetch and also PDOStatement (although the StatementInterface does not).

I've updated the patch with an additional test leveraging fetchObject with a classname (although no constructor).

Take #2 of the above, apologies although test ran, when trying to adding a record count check clearly didn't re-run the test - updated.

Lets test the patch with only the test in it.

Patch Rerolled for 9.1.x

@daffie, rerolled that patch against 9.1.x

@Neslee: When you have done a reroll, you are supposed to remove the tag "Needs reroll". Also the added test will fail. For that I will change the status to "Needs work". Thanks for the reroll.

I've trued to fix last patch

@aleevas: Could be so kind to also post a patch with the test and without the fix. The patch should fail, but then we know that the test is testing the fix.

The same patch as that from comment #17 only without the fix. This way we see if the testing is testing the fix.

The added test dos not test the bug fix.

StatementPrefetch is used by Sqlite not MySQL.

StatementPrefetch is used by Sqlite not MySQL.

You are completely right @johndevman. Lets say that I was doing something else and I am no good at multitasking. :)

The bug fix looks good and the added test looks good.
The pacth with only the added test shows that the added test is testing the bug fix.
For me it is RTBC.

+++ b/core/tests/Drupal/KernelTests/Core/Database/FetchTest.php
@@ -80,6 +80,24 @@ public function testQueryFetchClass() {
+      if ($this->assertTrue($result instanceof FakeRecord, 'Result is an object of class FakeRecord.')) {
+        $this->assertIdentical($result->name, 'John', '25 year old is John.');
+      }

The if here is strange. If the assertTrue fails then the test will halt so it's not necessary.

Also instead of if ($this->assertTrue($result instanceof FakeRecord, 'Result is an object of class FakeRecord.')) { this should be

  public function testQueryFetchObjectClass() {
    $records = 0;
    $query = $this->connection->query('SELECT name FROM {test} WHERE age = :age', [':age' => 25]);
    while ($result = $query->fetchObject(FakeRecord::class)) {
      $records += 1;
      $this->assertInstanceOf(FakeRecord::class, $result);
      $this->assertSame('John', $result->name, '25 year old is John.');
    }
    $this->assertSame(1, $records, 'There is only one record.');
  }

Yes this is copying other code in the class but it's all based on odd Simpletest conventions and not PHPUnit. $this->assertTrue returns NULL so amusingly none of these tests are testing what they think they are :) - can someone file a follow-up issue to fix all the tests in \Drupal\KernelTests\Core\Database\FetchTest() that are using the return value of assertTrue.

Addresses #23.

Follow up #3129074: Refactor assertions that use return values in conditionals

+++ b/core/tests/Drupal/KernelTests/Core/Database/FetchTest.php
@@ -80,6 +80,23 @@ public function testQueryFetchClass() {
+      $records += 1;
...
+    $this->assertCount(1, $records, 'There is only one record.');

$records is not a countable thing... in other test methods $records is an array not an integer. That's why in #23 I went for $this->assertSame(1, $records, 'There is only one record.');

#26 Yeah, sorry about that.

The code looks good.
All the points of @alexpott are addressed.
Back to RTBC.

Committed b2a1019 and pushed to 9.1.x. Thanks!

Going to ask release managers about backporting this bugfix.

Discussed with @catch and we agreed to backport this to 8.9.x