WebTestBase->setUp() broken for sqlite

Would this work so we don't have SQLite-specific exception handling?

if ($tables = db_find_tables($connection_info['default']['prefix']['default'] . '%');
 // Do stuff here...
}
else {
  $this->fail('Failed to find test tables to drop.');
}

1) I no longer know why I assigned the database settings manually, instead of just assigning the entire $connection_info['default'] as form values to submit to $edit. We should probably do that instead.

2) If db_find_tables() returns internal/native tables on sqlite, then we need to fix the database driver, instead of hacking around it in a single spot.

Sorry, but if the title of this issue is true, this is at least major.

2) If db_find_tables() returns internal/native tables on sqlite, then we need to fix the database driver, instead of hacking around it in a single spot.

There is another issue for prefixed database files failing to be cleaned-up on SQLite after testing. That is supposed to work (we have code in __destruct() of the SQLite implementation for that), but it seems to have been broken somewhere along the way.

Note that sqlite (and postgresql) will likely remain in this half-supported state until the new test infrastructure happens and we have test bots for them. This is a fact.

As far as I know, this is not a SQLite bug.

I think this is certainly a bug in Drupal's SQLite abstraction.

Everywhere in core we make the assumption that database table prefixes are database table prefixes (we don't have a generic namespacing interface that can be backed however it wants) SQLite violates that assumption and uses attached databases.

We could eventually change the db abstraction to add methods like "Create namespace" "Destroy namespace" or the like, but that's not the current case.

Because Simpltest (rightfully) assumes that it's dealing with table prefixes, it searches for and drops tables, but the sqlite abstraction layer fails to behave properly, and instead leaks its internal implementation details.

Actually, in order to allow for some progress to be made on setUp(), for which we have a solution in SimpleTest, I'm moving the tearDown() issue to another ticket (as it's possible the fix will be out in sqlite abstraction rather than simpletest.) New issue is here: #1814706: WebTestBase->tearDown() broken for sqlite

And here's my patch for this issue.

It matches sun's suggestion in #6

Thanks, simple and passes tests; looks good to me.

Looks good, and actually simplified things. Committed to 8.x. Thanks.

I'm not seeing the tearDown problem anymore either. I guess I'll close that issue.