Fix C-M-a in a C function finding the start of a macro preceding it.

Also amend some pertinent documentation.  This fixes bug #23818.

* lisp/progmodes/cc-engine.el (c-beginning-of-decl-1): Also check for a
virtual semicolon at a place where we check for other types of statement ends.

* lisp/progmodes/cc-vars.el (c-macro-nacmes-with-semicolon): Remove from the
doc string the bit saying that the variable is a prototype and liable to
change.

* doc/misc/cc-mode.texi (Macros with ;): Enhance, stating that configuring
macros with semicolon can prevent C-M-a missing the beginning of defun.

1 files changed, 4 insertions, 2 deletions

diff --git a/doc/misc/cc-mode.texi b/doc/misc/cc-mode.texi
index 82f8cbc2e3..f311ec8a3a 100644
--- a/doc/misc/cc-mode.texi
+++ b/doc/misc/cc-mode.texi
@@ -6727,9 +6727,11 @@ Macros which needn't (or mustn't) be followed by a semicolon when you
 invoke them, @dfn{macros with semicolons}, are very common.  These can
 cause @ccmode{} to parse the next line wrongly as a
 @code{statement-cont} (@pxref{Function Symbols}) and thus mis-indent
-it.
+it.  At the top level, a macro invocation before a defun start can
+cause, for example, @code{c-beginning-of-defun} (@kbd{C-M-a}) not to
+find the correct start of the current function.
 
-You can prevent this by specifying which macros have semicolons.  It
+You can prevent these by specifying which macros have semicolons.  It
 doesn't matter whether or not such a macro has a parameter list:
 
 @defopt c-macro-names-with-semicolon