Re: Somwhat subtle issues with raw string literals in C++

[Alan Mackenzie]

Hello again, Philipp,

On Tue, Jun 28, 2016 at 03:33:30PM +0000, Philipp Stephani wrote:
> Hi,

> there still seem to be some subtle issues with detection of raw string
> literals. Unfortunately they are hard to reproduce. One example that fails
> (for me) with a recent master build is:

> cd /tmp
> wget
> https://raw.githubusercontent.com/google/protobuf/ef7894e2dc6d287419e42a4fdc52cdfedd386d16/conformance/conformance_test.cc
> /path/to/emacs -Q +686 conformance_test.cc

> Around that line the fontification of the raw string literals is wrong
> (quote characters are treated as string terminators), in other parts of the
> file the fontification is correct. This typically happens with files that
> contain many large raw string literals that contain quote characters.

Again, thanks for the report.

The problem was a mishandling of a cache, with the result that the raw
string handling code mistakenly believed it was within a string when it
wasn't, at a critical point in the file.

Would you please try out the following patch and confirm that it fixes
the problem, or let me know what's still not working properly.

Thanks in advance!




diff -r 68a956fb5f55 cc-engine.el
--- a/cc-engine.el      Mon Jun 27 10:18:54 2016 +0000
+++ b/cc-engine.el      Wed Jun 29 13:43:27 2016 +0000
@@ -2292,6 +2292,12 @@
 ;; is reduced by buffer changes, and increased by invocations of
 ;; `c-parse-ps-state-below'.
 
+(defsubst c-truncate-semi-nonlit-pos-cache (pos)
+  ;; Truncate the upper bound of the cache `c-state-semi-nonlit-pos-cache' to
+  ;; POS, if it is higher than that position.
+  (setq c-state-semi-nonlit-pos-cache-limit
+       (min c-state-semi-nonlit-pos-cache-limit pos)))
+
 (defun c-state-semi-pp-to-literal (here &optional not-in-delimiter)
   ;; Do a parse-partial-sexp from a position in the buffer before HERE which
   ;; isn't in a literal, and return information about HERE, either:
@@ -2495,7 +2501,7 @@
       (let ((c c-state-semi-nonlit-pos-cache)
            elt state pos npos high-elt)
        ;; Trim the cache to take account of buffer changes.
-       (while (and c (> (c-ps-state-cache-pos (c-ps-state-cache-pos (car c)))
+       (while (and c (> (c-ps-state-cache-pos (car c))
                         c-state-semi-nonlit-pos-cache-limit))
          (setq c (cdr c)))
        (setq c-state-semi-nonlit-pos-cache c)
@@ -3488,8 +3494,7 @@
   ;; HERE.
   (if (<= here c-state-nonlit-pos-cache-limit)
       (setq c-state-nonlit-pos-cache-limit (1- here)))
-  (if (<= here c-state-semi-nonlit-pos-cache-limit)
-      (setq c-state-semi-nonlit-pos-cache-limit (1- here)))
+  (c-truncate-semi-nonlit-pos-cache here)
 
   ;; `c-state-cache':
   ;; Case 1: if `here' is in a literal containing point-min, everything
@@ -6082,19 +6085,32 @@
     (cond
      ((null open-paren-prop)
       ;; A terminated raw string
-      (if (search-forward (concat ")" id "\"") nil t)
-         (c-clear-char-property-with-value
-          (1+ open-paren) (match-beginning 0) 'syntax-table '(1))))
+      (when (search-forward (concat ")" id "\"") nil t)
+       (let* ((closing-paren (match-beginning 0))
+              (first-punctuation
+               (save-match-data
+                 (goto-char (1+ open-paren))
+                 (and (c-search-forward-char-property 'syntax-table '(1)
+                                                      closing-paren)
+                      (1- (point)))))
+              )
+         (when first-punctuation
+           (c-clear-char-property-with-value
+            first-punctuation (match-beginning 0) 'syntax-table '(1))
+           (c-truncate-semi-nonlit-pos-cache first-punctuation)
+           ))))
      ((or (and (equal open-paren-prop '(15)) (null bound))
          (equal open-paren-prop '(1)))
       ;; An unterminated raw string either not in a macro, or in a macro with
       ;; the open parenthesis right up against the end of macro
       (c-clear-char-property open-quote 'syntax-table)
+      (c-truncate-semi-nonlit-pos-cache open-quote)
       (c-clear-char-property open-paren 'syntax-table))
      (t
       ;; An unterminated string in a macro, with at least one char after the
       ;; open paren
       (c-clear-char-property open-quote 'syntax-table)
+      (c-truncate-semi-nonlit-pos-cache open-quote)
       (c-clear-char-property open-paren 'syntax-table)
       (let ((after-string-fence-pos
             (save-excursion
@@ -6194,9 +6210,11 @@
        (while (progn (skip-syntax-forward "^\"" end-string)
                      (< (point) end-string))
          (c-put-char-property (point) 'syntax-table '(1)) ; punctuation
+         (c-truncate-semi-nonlit-pos-cache (point))
          (forward-char))
        (goto-char after-quote))
     (c-put-char-property open-quote 'syntax-table '(1))             ; 
punctuation
+    (c-truncate-semi-nonlit-pos-cache open-quote)
     (c-put-char-property open-paren 'syntax-table '(15))     ; generic string
     (when bound
       ;; In a CPP construct, we try to apply a generic-string `syntax-table'
@@ -6223,9 +6241,12 @@
                       "\\(\\\\\n\\)*\\=")) ; 11
              (1+ open-paren) t))
          (if (match-beginning 10)
-             (c-put-char-property (match-beginning 10) 'syntax-table '(15))
+             (progn
+               (c-put-char-property (match-beginning 10) 'syntax-table '(15))
+               (c-truncate-semi-nonlit-pos-cache (match-beginning 10)))
            (c-put-char-property (match-beginning 5) 'syntax-table '(1))
-           (c-put-char-property (1+ (match-beginning 5)) 'syntax-table '(15)))
+           (c-put-char-property (1+ (match-beginning 5)) 'syntax-table '(15))
+           (c-truncate-semi-nonlit-pos-cache (1+ (match-beginning 5))))
        (c-put-char-property open-paren 'syntax-table '(1)))
       (goto-char bound))))
 


-- 
Alan Mackenzie (Nuremberg, Germany).