This code allows you to provide a code ref in the Object field in an Attribute
initialdata hash that returns an object that can be used as the ObjectType for
the Attribute:

@Attributes = ({
    Name        => 'SavedSearch',
    Description => 'New Tickets in SomeQueue',
    Object      => sub {
        my $GroupName = 'SomeQueue Group';
        my $group     = RT::Group->new( RT->SystemUser );

        my( $ret, $msg ) = $group->LoadUserDefinedGroup( $GroupName );
        die $msg unless $ret;

        return $group;
    },
    Content     => {
        Format =>  <<'        END_OF_FORMAT',
....
        END_OF_FORMAT
        Query   => "Status = 'new' AND Queue = 'SomeQueue'",
        OrderBy => 'id',
        Order   => 'DESC'
    },
});

Launchpad parsing objects to msgids which are not followed by empty
msgstr's in .pot files.

Old "message recorded" is confusing as it doesn't differentiate
between correspond and comment.

Passing $@ to loc() causes its value to not be examined until later,
when it is substituted into the string.  Between the ->require error and
then, the loc() machinery attempts to run for the first time, which
_also_ requires files, thus overwriting the original error in $@.

Store the error found in $@ away into a different variable, to ensure
that it is not lost before it is used.  Passing "$@" instead of $@ would
similarly suffice, but the clarity of the extra variable is preferrable.

Fixes I#22991.

As "GECOS" is what RT calls the field in a number of places, provide
that name in the error message.

Per I#30232.  This blacklist will be removed in a future release, likely
replaced by a dependency bump of DBIx::SearchBuilder for PostgreSQL
deployments.

The $LogoImageHeight and $LogoImageWidth configurations were only output
on the <img> tag if the stock BPS logo was in use.  Additionally,
however, those values were ignored by the browser, because every RT
style sets:

    div#logo a img {
         height: 100% !important;
         width: auto !important;
    }

...which overrides explicit height="..." and width=".." attributes.

Remove the configuration options which thus do not actually control
anything.

Fixes: I#26827

Pull request: #102

While 2a2198bc fixed the display of the HTML → text conversion for
text/plain embedded in an HTML template, it failed to address the
ckeditor, which suffered from a similar problem.

Notably, the ckeditor formatted <pre><div>...</div></pre> as
<pre></pre><div>...</div>; while the <div> maintained its style, all
whitespace had been compressed from its original value, leaving the
style pointless.

This table shows the previous states of this code, and its failures:

   MARKUP           |     MUA      |  HTML->TEXT   |   CKEDITOR
  ------------------+--------------+---------------+---------------
   <pre>            | doesn't wrap | GOOD          | GOOD
     75cdfc91       | long lines   |               |
  ------------------+--------------+---------------+---------------
   <div style>      | GOOD         | whitespace    | whitespace
     6b117af0       |              | not preserved | not preserved
  ------------------+--------------+---------------+---------------
   <pre><div style> | GOOD         | GOOD          | whitespace
     2a2198bc       |              |               | not preserved
  ------------------+--------------+---------------+---------------
   <pre style>      | GOOD         | GOOD          | GOOD
     (this commit)  |              |               |

Swap to using <pre style="...">...</pre>, which MUAs and ckeditor render
acceptably, and the stock HTML → text conversion preserves whitespace
in.

Fixes: I#30198