Clarify purpose of module aliases in ERC

* doc/misc/erc.texi: Mention that aliases should not be defined for
new modules.
* lisp/erc/erc-common.el (define-erc-module): Refactor slightly for
readability.
(erc-with-all-buffers-of-server): Redo doc string.
* lisp/erc/erc-pcomplete.el: Declare `completion' module's feature and
group as being `erc-pcomplete'.
* test/lisp/erc/erc-tests.el (erc--find-group--real): Assert group
lookup works for "normalized" module name `completion' of
`erc-pcomplete-mode'.

1 files changed, 8 insertions, 0 deletions

diff --git a/doc/misc/erc.texi b/doc/misc/erc.texi
index 52c7477c9dd..f877fb681fe 100644
--- a/doc/misc/erc.texi
+++ b/doc/misc/erc.texi
@@ -678,6 +678,14 @@ signals an error.  Users defining personal modules in an init file
 should @code{(provide 'erc-my-module)} somewhere to placate ERC.
 Dynamically generating modules on the fly is not supported.
 
+Some older built-in modules have a second name along with a second
+minor-mode toggle, which is just a function alias for its primary
+counterpart.  For practical reasons, ERC does not define a
+corresponding variable alias because contending with indirect
+variables complicates bookkeeping tasks, such as persisting module
+state across IRC sessions.  New modules should definitely avoid
+defining aliases without a good reason.
+
 Some packages have been known to autoload a module's definition
 instead of its minor-mode command, which severs the link between the
 library and the module.  This means that enabling the mode by invoking