Original Redmine Comment
Author Name: Wilson Snyder (@wsnyder)
Original Date: 2016-08-23T11:06:15Z

Thanks, this is a good suggestion.

Just so you know, the indentation code is pretty fragile and this would add more complication and perhaps performance issues, so it's unlikely to be fixed soon unless you take it on yourself.

Original Redmine Comment
Author Name: Kaushal Modi
Original Date: 2016-08-23T16:15:52Z

Is not trying to indent the lines ending in @\$@ at all a viable 'solution'?

Original Redmine Comment
Author Name: Wilson Snyder (@wsnyder)
Original Date: 2016-08-25T10:51:49Z

Is not trying to indent the lines ending in \$ at all a viable 'solution'?

A reasonable thing to be tried.

Original Redmine Comment
Author Name: Alex Reed
Original Date: 2016-08-25T16:39:50Z

Wilson Snyder wrote:

Is not trying to indent the lines ending in \$ at all a viable 'solution'?

A reasonable thing to be tried.

I tend to agree. I like the idea of semantically-aware indentation inside macros, but it's really not feasible. There's no guarantee that the `define is a complete (System)Verilog statement, therefore we cannot accurately indent the "Right Way" each and every time. Maybe abandoning all attempt at indentation is the best approach.

As a trivial example, consider a multi-line macro that contains quoted text (maybe it's part of a debug/error message). The user most definitely doesn't want their pretty message formatting perturbed by verilog-mode.

Original Redmine Comment
Author Name: Kaushal Modi
Original Date: 2016-08-25T17:41:31Z

Below tests fine in my personal config .. If that looks fine, I can start working on a patch for @verilog-indent-line-relative@ and @verilog-indent-line@.

(defvar modi/verilog-multi-line-define-line-cache nil
  "Variable set to non-nil if the current line is detected as any but the
last line of a multi-line `define such as:

  `define foo(ARG) \          <- non-nil
     begin \                   <- non-nil
       $display(\"Bar\"); \    <- non-nil
       $display(\"Baz\"); \    <- non-nil
     end                       <- nil
 ")

(defun modi/verilog-selective-indent (&rest args)
  "Return non-nil if point is on certain types of lines.

Non-nil return will happen when the current line is part of a multi-line `define like:
     `define foo(ARG) \
       begin \
         $display(\"Bar\"); \
         $display(\"Baz\"); \
       end

This function is used to tweak the `verilog-mode' indentation functions. "
  (save-excursion
     (beginning-of-line)
     (let* ((is-in-multi-line-define (looking-at "^.*\\\\$")) ; \ at EOL
            (do-not-run-orig-fn (or is-in-multi-line-define
                                    modi/verilog-multi-line-define-line-cache)))
       ;; Cache the current value of `is-in-multi-line-define'
       (setq modi/verilog-multi-line-define-line-cache is-in-multi-line-define)
       do-not-run-orig-fn)))
;; Advise the indentation behavior of `indent-region' done using `C-M-\'
(advice-add 'verilog-indent-line-relative :before-until #'modi/verilog-selective-indent)
;; Advise the indentation done by hitting `TAB'
(advice-add 'verilog-indent-line :before-until #'modi/verilog-selective-indent)

Hi @kaushalmodi , @wsnyder

I have written some code that supports ignoring indentation of multiline macros and custom regexps conditionally through defcustoms (5b9bc2a). It works on GNU Emacs 28.1.

This would enable ignoring indentation for outshine comments (#885), custom "keep indent" markers (#922), or any other custom regexp.

Since this does not rely on advice-add, could this be merged into master?

Thanks!

Can you please convert to a pull request? With that not sure we need ignore-regexp.

Also should we say verilog-indent-ignore-multiline-defines instead of -macros?

I don't think we should be changing indentation of any comments that are outside `defines, not sure if you are.

The verilog-indent-ignore-regexp variable is meant to be used for different things than multiline defines. It defaults to nil so there is no change unless enabled explicitly.

In #885 @kaushalmodi explains why it would be nice to be able to ignore indentation on some specific lines. I also use outshine along with modi's advice function and it works really well for buffer navigation. However, since his solution seems very user-specific the use of a variable with a custom regexp would be much more generic.

If this variable is set to "// \\*" it has the same effect as modi's adviced function. If set to ".*//\\." it will match expressions mentioned in #922. Most likely this function will not be of much use to most people, but it could be very helpful for some particular cases.

Implemented in #1778