Fix errors uncovered using aggressive watchdog assertion

+++ includes/locale.inc	6 Dec 2009 07:51:37 -0000
@@ -242,6 +242,7 @@ function locale_languages_edit_form($for
+    drupal_exit();

Are you sure you want to call drupal_exit() and not 'return'? I'd have to dig a bit deeper but it could affect some callbacks not being called.

+++ includes/locale.inc	6 Dec 2009 07:51:37 -0000
@@ -1695,7 +1697,7 @@ function _locale_import_po($file, $langc
-    watchdog('locale', $skip_message, NULL, WATCHDOG_WARNING);
+    watchdog('locale', '%file: Disallowed HTML in @count string(s)', array('%file' => $file->filename, '@count' => $skips), WATCHDOG_WARNING);
   }

It might be slightly better to move %file to appear later in the string. We only see x characters on the watchdog overview page and I'm worried that we'd see the filename only. We should lead with the part of the message that provides context.

Committed to CVS HEAD. Thanks!

I reviewed #10 with sun in IRC:

I asked about:

+    $this->drupalPost(NULL, $edit, t('Save configuration'));

This means "post back to the current page." I'm sure if I had the drupalPost docs in front of me rather than reading a patch file, I would know that. ;)

I was also concerned about how often this patch adds lines like this:

+    $this->assertResponse(404);
...
+    $this->assertNoResponse(404);

...and how to explain to test authors when they need to do so.

sun explained that you do these explicitly when your test depends on a 404/403 response; for example, while testing that the search block appears on the 404 page.

Committed #10 to HEAD, marking back to active.

Er, am I going crazy or does this assertNoResponse() method not exist in simpletest? All the tests that use this method fail with:

Fatal error: Call to undefined method PathTestCase::assertNoResponse() in httpdocs/modules/path/path.test on line 67

Isn't the method supposed to be in drupal_web_test_case.php? http://drupalcode.org/viewvc/drupal/drupal/modules/simpletest/drupal_web...

agreed! that was really weird. glad i wasn't crazy.

#13 fixes the fatals in HEAD, let's get it committed then set back to active.

Yep. Slightly more quick-commit friendly issue-title too ;)

Committed to CVS HEAD. Thanks.

Back to the remainder of #10.

Anything but normal.

It looks like #13 wasn't committed. Instead the patch for #459786: countQuery() should also remove all_fields statements was committed twice by mistake. The last commit message had a reference to this issue.

Confirmed this absolutely broke core testing.

Confirmed #21 that #13 wasn't committed. Changing the title back with some emphasis.

Committed to CVS HEAD. For real now. The network at DIDW is really unreliable so sometimes things hang for minutes.

Changing title to something a little more descriptive :D