bug#12274: 24.2; awk-mode indentation failure

bug#12274: 24.2; awk-mode indentation failure

[Leo]

1. Emacs -q and change to a new buffer
2. M-x awk-mode
3. Type: NR >= 3
4. Type return and then tab to indent

The newline is indented.

AWK mode fails to recognise NR >= 3 as a pattern-action pair where
action is omitted. Statements in awk need no ; to terminate.

Leo

bug#13541: 24.2.92; awk-mode: wrong font locking regexp literals

[Leo Liu]

[-- Attachment #1: Type: text/plain, Size: 214 bytes --]

In an awk buffer having the following text:

#--BEGIN--
NF { /xyz/ }

NF {
    /xyz/
}
#--END--

I have the second regexp properly font-locked but not the first one.

(tested in GNU Emacs 24.2.92.1 of 2013-01-13)


[-- Attachment #2: awk-mode-bug.png --]
[-- Type: image/png, Size: 6820 bytes --]

[-- Attachment #3: Type: text/plain, Size: 5 bytes --]


Leo

bug#13541: 24.2.92; awk-mode: wrong font locking regexp literals

[Alan Mackenzie]

Hi, Glenn,

On Thu, Jan 24, 2013 at 01:28:33PM -0500, Glenn Morris wrote:
> Leo Liu wrote:

> > In an awk buffer having the following text:

> > #--BEGIN--
> > NF { /xyz/ }

> > NF {
> >     /xyz/
> > }
> > #--END--

> > I have the second regexp properly font-locked but not the first one.

> Do you have an example of an actual useful awk script showing the issue,
> because this one seems like a pointless no-op?

This is a real bug, perhaps not a difficult one.  "/regexp/" is an
expression with value 1 iff the current input line matches the regexp.
So a line like

    NF { print /xyz/ }

is perfectly legitimate, printing 1 if there's an "xyz" on the line.

I'm looking at this bug at the moment.

-- 
Alan Mackenzie (Nuremberg, Germany).

bug#13541: 24.2.92; awk-mode: wrong font locking regexp literals

[Leo Liu]

On 2013-01-25 06:16 +0800, Alan Mackenzie wrote:
> This is a real bug, perhaps not a difficult one.  "/regexp/" is an
> expression with value 1 iff the current input line matches the regexp.
> So a line like
>
>     NF { print /xyz/ }
>
> is perfectly legitimate, printing 1 if there's an "xyz" on the line.
>
> I'm looking at this bug at the moment.

Thanks to all for chiming in.

Alan, I also have another seemingly buglet about indentation.

Every line after a pattern-action pair like the following one (where
action is omitted) is indented to column 4, i.e. it doesn't recognise a
newline terminates a pattern.

$0 == "Emacs"
    |
    all following lines indented here

(this might be regression, I seem to recall reporting something along
these lines some while ago.)

Leo

bug#12274: 24.2; awk-mode indentation failure

[Alan Mackenzie]

Hello, Leo.

On Fri, Jan 25, 2013 at 09:20:19AM +0800, Leo Liu wrote:

[ .... ]

> Alan, I also have another seemingly buglet about indentation.

As Glenn remarked, this is bug #12274.

> Every line after a pattern-action pair like the following one (where
> action is omitted) is indented to column 4, i.e. it doesn't recognise a
> newline terminates a pattern.

> $0 == "Emacs"
>     |
>     all following lines indented here

> (this might be regression, I seem to recall reporting something along
> these lines some while ago.)

No, not a regression, rather a bug which has been there since 4004 BC.
It's actually the "=" sign which triggers it, confusing the parsing
algortihm into thinking it's a C initialisation statement.

The solution is to move the pertinent AWK parsing clause earlier on in
the enclosing cond form.

Glenn, this is not a regression.  Should I nevertheless commit it to the
emacs-24 branch?

Here's the patch:


diff -r 0d641a4d3e7c cc-engine.el
--- a/cc-engine.el	Wed Jan 23 18:17:40 2013 +0000
+++ b/cc-engine.el	Fri Jan 25 08:27:12 2013 +0000
@@ -9880,6 +9880,18 @@
 	    ;; contains any class offset
 	    )))
 
+	 ;; CASE 5P: AWK pattern or function or continuation
+	 ;; thereof.
+	 ((c-major-mode-is 'awk-mode)
+	  (setq placeholder (point))
+	  (c-add-stmt-syntax
+	   (if (and (eq (c-beginning-of-statement-1) 'same)
+		    (/= (point) placeholder))
+	       'topmost-intro-cont
+	     'topmost-intro)
+	   nil nil
+	   containing-sexp paren-state))
+
 	 ;; CASE 5D: this could be a top-level initialization, a
 	 ;; member init list continuation, or a template argument
 	 ;; list continuation.
@@ -10039,18 +10051,6 @@
 	      (goto-char (point-min)))
 	  (c-add-syntax 'objc-method-intro (c-point 'boi)))
 
-	 ;; CASE 5P: AWK pattern or function or continuation
-	 ;; thereof.
-	 ((c-major-mode-is 'awk-mode)
-	  (setq placeholder (point))
-	  (c-add-stmt-syntax
-	   (if (and (eq (c-beginning-of-statement-1) 'same)
-		    (/= (point) placeholder))
-	       'topmost-intro-cont
-	     'topmost-intro)
-	   nil nil
-	   containing-sexp paren-state))
-
 	 ;; CASE 5N: At a variable declaration that follows a class
 	 ;; definition or some other block declaration that doesn't
 	 ;; end at the closing '}'.  C.f. case 5D.5.

> Leo

-- 
Alan Mackenzie (Nuremberg, Germany).

bug#12274: 24.2; awk-mode indentation failure

[Stefan Monnier]

> Glenn, this is not a regression.  Should I nevertheless commit it to the
> emacs-24 branch?

No, this should go to trunk,


        Stefan

bug#12274: 24.2; awk-mode indentation failure

[Glenn Morris]

Alan Mackenzie wrote:

> Glenn, this is not a regression.  Should I nevertheless commit it to the
> emacs-24 branch?

No, trunk please.

bug#12274: 24.2; awk-mode indentation failure

[Alan Mackenzie]

Bug fixed.

-- 
Alan Mackenzie (Nuremberg, Germany).