Robert Spier wrote:
The hash is only used to check for validity of the hook name. If it
helps us, we could split out @hooks and %hooks.
I only said I /thought/ it would work if we could control the order of hook_*...
<later>
Well, it turns out that it mostly works (in that it doesn't recurse), and I had
to make C<sub register> into C<sub init> so that the initialization happens
before the C<sub hook_logging>. But, then I get lines like this:
Subroutine plugin_name redefined at (eval 24) line 8.
Subroutine hook_name redefined at (eval 24) line 9.
Subroutine init redefined at ./plugins/logging/adaptive line 5.
Subroutine hook_logging redefined at ./plugins/logging/adaptive line 38.
Subroutine hook_deny redefined at ./plugins/logging/adaptive line 65.
Subroutine hook_reset_transaction redefined at ./plugins/logging/adaptive line
70.
If I tweak hook_logging to not expect the initialization to have happened when
it is registered (and go back to using register), I don't get those lines
(weird). So there may be something odd about the init() handling still. I've
re-reverted the changes to plugins/logging/* and commited a change to
Qpsmtpd::Plugin to track hooks as array and hash.
John
