bug#33163: Adding a " in c++-mode fills the whole file with red

bug#33163: Adding a " in c++-mode fills the whole file with red

[Konstantin Kharlamov]

This long-standing problem appears occasionally, and only disappears if you kill the buffer completely, then reopen the file. I never got to report it because I never could find steps to reproduce it… until now :) Nvm the non-sense code, I reduced testcase as far as I managed.

# Steps to reproduce:

1. Save the following text in `test.cpp`
     int main() {
             puts(It's");
         auto foo = []()-> void {  };
             puts("It's");
     }
2. Open `emacs -Q ./test.cpp`
3. Now, the text lacks a quote in the first `puts(It's");`. Add it so it looks like `puts("It's");`. Essentially, if you'd compile it with g++, it lacks any syntax error besides `puts` being undeclared.

# Expected

The line with `auto foo…` is not red.

# Actual

The line with `auto foo…` (in particular) becomes red.

# Additional information

Emacs version GNU Emacs 27.0.50 (build 1, x86_64-pc-linux-gnu, GTK+ Version 3.22.30) of 2018-10-01, build with -flto.

bug#33163: Adding a " in c++-mode fills the whole file with red

[Alan Mackenzie]

In article <mailman.2795.1540562775.1284.bug-gnu-emacs@gnu.org> you wrote:
> This long-standing problem appears occasionally, and only disappears if
> you kill the buffer completely, then reopen the file. I never got to
> report it because I never could find steps to reproduce it… until now
> :) Nvm the non-sense code, I reduced testcase as far as I managed.

> # Steps to reproduce:

> 1. Save the following text in `test.cpp`
>     int main() {
>             puts(It's");
>         auto foo = []()-> void {  };
>             puts("It's");
>     }
> 2. Open `emacs -Q ./test.cpp`
> 3. Now, the text lacks a quote in the first `puts(It's");`. Add it so
> it looks like `puts("It's");`. Essentially, if you'd compile it with
> g++, it lacks any syntax error besides `puts` being undeclared.

> # Expected

> The line with `auto foo…` is not red.

> # Actual

> The line with `auto foo…` (in particular) becomes red.

Thanks for taking the trouble to report this bug.  Thanks even more for
condensing it down to a minimal test case.

The following patch should fix it.  Would you please apply the patch,
recompile .../lisp/progmodes/cc-mode.el, load the new file in to your
Emacs (or restart Emacs), then try it out on your real C++ code.  Then
please let me know whether it's working or not, and if not, how it's
going wrong.  Thanks!

> # Additional information

> Emacs version GNU Emacs 27.0.50 (build 1, x86_64-pc-linux-gnu, GTK+
> Version 3.22.30) of 2018-10-01, build with -flto.



diff -r 2f7029fa229d cc-mode.el
--- a/cc-mode.el	Sun Oct 14 20:01:07 2018 +0000
+++ b/cc-mode.el	Sun Oct 28 10:20:38 2018 +0000
@@ -1216,21 +1216,21 @@
       (if (eq beg-literal-type 'string)
 	  (setq c-new-BEG (min (car beg-limits) c-new-BEG))))
 
-     ((< c-new-END (point-max))
-      (goto-char (1+ c-new-END))	; might be a newline.
+     ((< end (point-max))
+      (goto-char (1+ end))	; might be a newline.
       ;; In the following regexp, the initial \n caters for a newline getting
       ;; joined to a preceding \ by the removal of what comes between.
       (re-search-forward "[\n\r]?\\(\\\\\\(.\\|\n\\|\r\\)\\|[^\\\n\r]\\)*"
 			 nil t)
       ;; We're at an EOLL or point-max.
-      (setq c-new-END (min (1+ (point)) (point-max)))
-      (goto-char c-new-END)
-      (if (equal (c-get-char-property (1- (point)) 'syntax-table) '(15))
-	  (if (memq (char-before) '(?\n ?\r))
+      (setq c-new-END (max c-new-END (min (1+ (point)) (point-max))))
+      (if (equal (c-get-char-property (point) 'syntax-table) '(15))
+	  (if (memq (char-after) '(?\n ?\r))
 	      ;; Normally terminated invalid string.
-	      (progn
+	      (let ((eoll-1 (point)))
+		(forward-char)
 		(backward-sexp)
-		(c-clear-char-property (1- c-new-END) 'syntax-table)
+		(c-clear-char-property eoll-1 'syntax-table)
 		(c-clear-char-property (point) 'syntax-table))
 	    ;; Opening " at EOB.
 	    (c-clear-char-property (1- (point)) 'syntax-table))
@@ -1238,7 +1238,7 @@
 	    ;; Opening " on last line of text (without EOL).
 	    (c-clear-char-property (point) 'syntax-table))))
 
-     (t (goto-char c-new-END)
+     (t (goto-char end)			; point-max
 	(if (c-search-backward-char-property 'syntax-table '(15) c-new-BEG)
 	    (c-clear-char-property (point) 'syntax-table))))
 
@@ -1327,9 +1327,9 @@
 	     (while (progn
 		      (setq s (parse-partial-sexp (point) c-new-END nil
 						  nil s 'syntax-table))
-		      (and (not (nth 3 s))
-			   (< (point) c-new-END)
-			   (not (memq (char-before) c-string-delims)))))
+                      (and (< (point) c-new-END)
+                           (or (not (nth 3 s))
+                               (not (memq (char-before) c-string-delims))))))
 	     ;; We're at the start of a string.
 	     (memq (char-before) c-string-delims)))
 	(if (c-unescaped-nls-in-string-p (1- (point)))


-- 
Alan Mackenzie (Nuremberg, Germany).

bug#33163: Adding a " in c++-mode fills the whole file with red

[Konstantin Kharlamov]

On 10/28/18 1:49 PM, Alan Mackenzie wrote:
> In article <mailman.2795.1540562775.1284.bug-gnu-emacs@gnu.org> you wrote:
>> This long-standing problem appears occasionally, and only disappears if
>> you kill the buffer completely, then reopen the file. I never got to
>> report it because I never could find steps to reproduce it… until now
>> :) Nvm the non-sense code, I reduced testcase as far as I managed.
> 
>> # Steps to reproduce:
> 
>> 1. Save the following text in `test.cpp`
>>      int main() {
>>              puts(It's");
>>          auto foo = []()-> void {  };
>>              puts("It's");
>>      }
>> 2. Open `emacs -Q ./test.cpp`
>> 3. Now, the text lacks a quote in the first `puts(It's");`. Add it so
>> it looks like `puts("It's");`. Essentially, if you'd compile it with
>> g++, it lacks any syntax error besides `puts` being undeclared.
> 
>> # Expected
> 
>> The line with `auto foo…` is not red.
> 
>> # Actual
> 
>> The line with `auto foo…` (in particular) becomes red.
> 
> Thanks for taking the trouble to report this bug.  Thanks even more for
> condensing it down to a minimal test case.
> 
> The following patch should fix it.  Would you please apply the patch,
> recompile .../lisp/progmodes/cc-mode.el, load the new file in to your
> Emacs (or restart Emacs), then try it out on your real C++ code.  Then
> please let me know whether it's working or not, and if not, how it's
> going wrong.  Thanks!

Thank you for the quick patch! I tried to do some editing and 
navigating, and all seems well. More real testing I can only do on 
Monday, so I guess I'll reply somewhere on Monday evening.

> 
>> # Additional information
> 
>> Emacs version GNU Emacs 27.0.50 (build 1, x86_64-pc-linux-gnu, GTK+
>> Version 3.22.30) of 2018-10-01, build with -flto.
> 
> 
> 
> diff -r 2f7029fa229d cc-mode.el
> --- a/cc-mode.el	Sun Oct 14 20:01:07 2018 +0000
> +++ b/cc-mode.el	Sun Oct 28 10:20:38 2018 +0000
> @@ -1216,21 +1216,21 @@
>         (if (eq beg-literal-type 'string)
>   	  (setq c-new-BEG (min (car beg-limits) c-new-BEG))))
>   
> -     ((< c-new-END (point-max))
> -      (goto-char (1+ c-new-END))	; might be a newline.
> +     ((< end (point-max))
> +      (goto-char (1+ end))	; might be a newline.
>         ;; In the following regexp, the initial \n caters for a newline getting
>         ;; joined to a preceding \ by the removal of what comes between.
>         (re-search-forward "[\n\r]?\\(\\\\\\(.\\|\n\\|\r\\)\\|[^\\\n\r]\\)*"
>   			 nil t)
>         ;; We're at an EOLL or point-max.
> -      (setq c-new-END (min (1+ (point)) (point-max)))
> -      (goto-char c-new-END)
> -      (if (equal (c-get-char-property (1- (point)) 'syntax-table) '(15))
> -	  (if (memq (char-before) '(?\n ?\r))
> +      (setq c-new-END (max c-new-END (min (1+ (point)) (point-max))))
> +      (if (equal (c-get-char-property (point) 'syntax-table) '(15))
> +	  (if (memq (char-after) '(?\n ?\r))
>   	      ;; Normally terminated invalid string.
> -	      (progn
> +	      (let ((eoll-1 (point)))
> +		(forward-char)
>   		(backward-sexp)
> -		(c-clear-char-property (1- c-new-END) 'syntax-table)
> +		(c-clear-char-property eoll-1 'syntax-table)
>   		(c-clear-char-property (point) 'syntax-table))
>   	    ;; Opening " at EOB.
>   	    (c-clear-char-property (1- (point)) 'syntax-table))
> @@ -1238,7 +1238,7 @@
>   	    ;; Opening " on last line of text (without EOL).
>   	    (c-clear-char-property (point) 'syntax-table))))
>   
> -     (t (goto-char c-new-END)
> +     (t (goto-char end)			; point-max
>   	(if (c-search-backward-char-property 'syntax-table '(15) c-new-BEG)
>   	    (c-clear-char-property (point) 'syntax-table))))
>   
> @@ -1327,9 +1327,9 @@
>   	     (while (progn
>   		      (setq s (parse-partial-sexp (point) c-new-END nil
>   						  nil s 'syntax-table))
> -		      (and (not (nth 3 s))
> -			   (< (point) c-new-END)
> -			   (not (memq (char-before) c-string-delims)))))
> +                      (and (< (point) c-new-END)
> +                           (or (not (nth 3 s))
> +                               (not (memq (char-before) c-string-delims))))))
>   	     ;; We're at the start of a string.
>   	     (memq (char-before) c-string-delims)))
>   	(if (c-unescaped-nls-in-string-p (1- (point)))
> 
> 

bug#33163: Adding a " in c++-mode fills the whole file with red

[Konstantin Kharlamov]

Thank you, so, I've used the patch today for working with a code, and 
everything seemed to be well :)

Tested-by: Konstantin Kharlamov <hi-angel@yandex.ru>

On 10/28/18 2:21 PM, Konstantin Kharlamov wrote:
> On 10/28/18 1:49 PM, Alan Mackenzie wrote:
>> In article <mailman.2795.1540562775.1284.bug-gnu-emacs@gnu.org> you 
>> wrote:
>>> This long-standing problem appears occasionally, and only disappears if
>>> you kill the buffer completely, then reopen the file. I never got to
>>> report it because I never could find steps to reproduce it… until now
>>> :) Nvm the non-sense code, I reduced testcase as far as I managed.
>>
>>> # Steps to reproduce:
>>
>>> 1. Save the following text in `test.cpp`
>>>      int main() {
>>>              puts(It's");
>>>          auto foo = []()-> void {  };
>>>              puts("It's");
>>>      }
>>> 2. Open `emacs -Q ./test.cpp`
>>> 3. Now, the text lacks a quote in the first `puts(It's");`. Add it so
>>> it looks like `puts("It's");`. Essentially, if you'd compile it with
>>> g++, it lacks any syntax error besides `puts` being undeclared.
>>
>>> # Expected
>>
>>> The line with `auto foo…` is not red.
>>
>>> # Actual
>>
>>> The line with `auto foo…` (in particular) becomes red.
>>
>> Thanks for taking the trouble to report this bug.  Thanks even more for
>> condensing it down to a minimal test case.
>>
>> The following patch should fix it.  Would you please apply the patch,
>> recompile .../lisp/progmodes/cc-mode.el, load the new file in to your
>> Emacs (or restart Emacs), then try it out on your real C++ code.  Then
>> please let me know whether it's working or not, and if not, how it's
>> going wrong.  Thanks!
> 
> Thank you for the quick patch! I tried to do some editing and 
> navigating, and all seems well. More real testing I can only do on 
> Monday, so I guess I'll reply somewhere on Monday evening.
> 
>>
>>> # Additional information
>>
>>> Emacs version GNU Emacs 27.0.50 (build 1, x86_64-pc-linux-gnu, GTK+
>>> Version 3.22.30) of 2018-10-01, build with -flto.
>>
>>
>>
>> diff -r 2f7029fa229d cc-mode.el
>> --- a/cc-mode.el    Sun Oct 14 20:01:07 2018 +0000
>> +++ b/cc-mode.el    Sun Oct 28 10:20:38 2018 +0000
>> @@ -1216,21 +1216,21 @@
>>         (if (eq beg-literal-type 'string)
>>         (setq c-new-BEG (min (car beg-limits) c-new-BEG))))
>> -     ((< c-new-END (point-max))
>> -      (goto-char (1+ c-new-END))    ; might be a newline.
>> +     ((< end (point-max))
>> +      (goto-char (1+ end))    ; might be a newline.
>>         ;; In the following regexp, the initial \n caters for a 
>> newline getting
>>         ;; joined to a preceding \ by the removal of what comes between.
>>         (re-search-forward 
>> "[\n\r]?\\(\\\\\\(.\\|\n\\|\r\\)\\|[^\\\n\r]\\)*"
>>                nil t)
>>         ;; We're at an EOLL or point-max.
>> -      (setq c-new-END (min (1+ (point)) (point-max)))
>> -      (goto-char c-new-END)
>> -      (if (equal (c-get-char-property (1- (point)) 'syntax-table) '(15))
>> -      (if (memq (char-before) '(?\n ?\r))
>> +      (setq c-new-END (max c-new-END (min (1+ (point)) (point-max))))
>> +      (if (equal (c-get-char-property (point) 'syntax-table) '(15))
>> +      (if (memq (char-after) '(?\n ?\r))
>>             ;; Normally terminated invalid string.
>> -          (progn
>> +          (let ((eoll-1 (point)))
>> +        (forward-char)
>>           (backward-sexp)
>> -        (c-clear-char-property (1- c-new-END) 'syntax-table)
>> +        (c-clear-char-property eoll-1 'syntax-table)
>>           (c-clear-char-property (point) 'syntax-table))
>>           ;; Opening " at EOB.
>>           (c-clear-char-property (1- (point)) 'syntax-table))
>> @@ -1238,7 +1238,7 @@
>>           ;; Opening " on last line of text (without EOL).
>>           (c-clear-char-property (point) 'syntax-table))))
>> -     (t (goto-char c-new-END)
>> +     (t (goto-char end)            ; point-max
>>       (if (c-search-backward-char-property 'syntax-table '(15) c-new-BEG)
>>           (c-clear-char-property (point) 'syntax-table))))
>> @@ -1327,9 +1327,9 @@
>>            (while (progn
>>                 (setq s (parse-partial-sexp (point) c-new-END nil
>>                             nil s 'syntax-table))
>> -              (and (not (nth 3 s))
>> -               (< (point) c-new-END)
>> -               (not (memq (char-before) c-string-delims)))))
>> +                      (and (< (point) c-new-END)
>> +                           (or (not (nth 3 s))
>> +                               (not (memq (char-before) 
>> c-string-delims))))))
>>            ;; We're at the start of a string.
>>            (memq (char-before) c-string-delims)))
>>       (if (c-unescaped-nls-in-string-p (1- (point)))
>>
>>

bug#33163: Adding a " in c++-mode fills the whole file with red

[Alan Mackenzie]

Hello, Konstantin

On Mon, Oct 29, 2018 at 21:46:50 +0300, Konstantin Kharlamov wrote:
> Thank you, so, I've used the patch today for working with a code, and 
> everything seemed to be well :)

> Tested-by: Konstantin Kharlamov <hi-angel@yandex.ru>

Thanks for the testing.  I've committed the patch to all the usual
places, and I'm closing the bug.

-- 
Alan Mackenzie (Nuremberg, Germany).

bug#33163: Adding a " in c++-mode fills the whole file with red

[Konstantin Kharlamov]

FTR: since then I've encountered a likewise problem a few times more,
except now it only happens with lines `#include "file.h"`. I haven't
manage yet to get a minimal reproducible example, but I thought it's
worth mentioning here, just in case you might have an offhand insight.

On 30.10.2018 15:01, Alan Mackenzie wrote:
> Hello, Konstantin
> 
> On Mon, Oct 29, 2018 at 21:46:50 +0300, Konstantin Kharlamov wrote:
>> Thank you, so, I've used the patch today for working with a code, and
>> everything seemed to be well :)
> 
>> Tested-by: Konstantin Kharlamov <hi-angel@yandex.ru>
> 
> Thanks for the testing.  I've committed the patch to all the usual
> places, and I'm closing the bug.
>