Blob Blame History Raw
From 8f907e8731b5eb4aca840a1979e76eed1b5c8175 Mon Sep 17 00:00:00 2001
From: ph10 <ph10@2f5784b3-3f2a-0410-8824-cb99058d5e15>
Date: Tue, 7 Apr 2015 15:52:11 +0000
Subject: [PATCH] Fix pcregrep loop when \K is used in a lookbehind assertion.
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Upstream commit ported to 8.36:

commit b4332d7dd831b3547b3f541495de4a79554e538e
Author: ph10 <ph10@2f5784b3-3f2a-0410-8824-cb99058d5e15>
Date:   Tue Apr 7 15:52:11 2015 +0000

    Fix pcregrep loop when \K is used in a lookbehind assertion.

    git-svn-id: svn://vcs.exim.org/pcre/code/trunk@1543 2f5784b3-3f2a-0410-8824-cb99058d5e15

Signed-off-by: Petr Písař <ppisar@redhat.com>
---
 RunGrepTest         |   5 +++
 pcregrep.c          | 109 ++++++++++++++++++++++++++++++++++------------------
 testdata/grepoutput |   8 ++++
 3 files changed, 85 insertions(+), 37 deletions(-)

diff --git a/RunGrepTest b/RunGrepTest
index f1b0348..766278b 100755
--- a/RunGrepTest
+++ b/RunGrepTest
@@ -506,6 +506,11 @@ echo "---------------------------- Test 106 -----------------------------" >>tes
 (cd $srcdir; echo "a" | $valgrind $pcregrep -M "|a" ) >>testtrygrep 2>&1
 echo "RC=$?" >>testtrygrep
 
+echo "---------------------------- Test 107 -----------------------------" >>testtrygrep
+echo "a" >testtemp1grep
+echo "aaaaa" >>testtemp1grep
+(cd $srcdir; $valgrind $pcregrep  --line-offsets '(?<=\Ka)' testtemp1grep) >>testtrygrep 2>&1
+echo "RC=$?" >>testtrygrep
 
 # Now compare the results.
 
diff --git a/pcregrep.c b/pcregrep.c
index 4f7fa38..b1af129 100644
--- a/pcregrep.c
+++ b/pcregrep.c
@@ -1582,11 +1582,14 @@ while (ptr < endptr)
   int endlinelength;
   int mrc = 0;
   int startoffset = 0;
+  int prevoffsets[2]; 
   unsigned int options = 0;
   BOOL match;
   char *matchptr = ptr;
   char *t = ptr;
   size_t length, linelength;
+  
+  prevoffsets[0] = prevoffsets[1] = -1; 
 
   /* At this point, ptr is at the start of a line. We need to find the length
   of the subject string to pass to pcre_exec(). In multiline mode, it is the
@@ -1729,55 +1732,86 @@ while (ptr < endptr)
       {
       if (!invert)
         {
-        if (printname != NULL) fprintf(stdout, "%s:", printname);
-        if (number) fprintf(stdout, "%d:", linenumber);
-
-        /* Handle --line-offsets */
-
-        if (line_offsets)
-          fprintf(stdout, "%d,%d\n", (int)(matchptr + offsets[0] - ptr),
-            offsets[1] - offsets[0]);
-
-        /* Handle --file-offsets */
-
-        else if (file_offsets)
-          fprintf(stdout, "%d,%d\n",
-            (int)(filepos + matchptr + offsets[0] - ptr),
-            offsets[1] - offsets[0]);
-
-        /* Handle --only-matching, which may occur many times */
-
-        else
+        int oldstartoffset = startoffset;
+        
+        /* It is possible, when a lookbehind assertion contains \K, for the 
+        same string to be found again. The code below advances startoffset, but 
+        until it is past the "bumpalong" offset that gave the match, the same
+        substring will be returned. The PCRE1 library does not return the
+        bumpalong offset, so all we can do is ignore repeated strings. (PCRE2
+        does this better.) */
+         
+        if (prevoffsets[0] != offsets[0] || prevoffsets[1] != offsets[1])
           {
-          BOOL printed = FALSE;
-          omstr *om;
-
-          for (om = only_matching; om != NULL; om = om->next)
+          prevoffsets[0] = offsets[0];
+          prevoffsets[1] = offsets[1]; 
+            
+          if (printname != NULL) fprintf(stdout, "%s:", printname);
+          if (number) fprintf(stdout, "%d:", linenumber);
+          
+          /* Handle --line-offsets */
+          
+          if (line_offsets)
+            fprintf(stdout, "%d,%d\n", (int)(matchptr + offsets[0] - ptr),
+              offsets[1] - offsets[0]);
+          
+          /* Handle --file-offsets */
+          
+          else if (file_offsets)
+            fprintf(stdout, "%d,%d\n",
+              (int)(filepos + matchptr + offsets[0] - ptr),
+              offsets[1] - offsets[0]);
+          
+          /* Handle --only-matching, which may occur many times */
+          
+          else
             {
-            int n = om->groupnum;
-            if (n < mrc)
+            BOOL printed = FALSE;
+            omstr *om;
+          
+            for (om = only_matching; om != NULL; om = om->next)
               {
-              int plen = offsets[2*n + 1] - offsets[2*n];
-              if (plen > 0)
+              int n = om->groupnum;
+              if (n < mrc)
                 {
-                if (printed) fprintf(stdout, "%s", om_separator);
-                if (do_colour) fprintf(stdout, "%c[%sm", 0x1b, colour_string);
-                FWRITE(matchptr + offsets[n*2], 1, plen, stdout);
-                if (do_colour) fprintf(stdout, "%c[00m", 0x1b);
-                printed = TRUE;
+                int plen = offsets[2*n + 1] - offsets[2*n];
+                if (plen > 0)
+                  {
+                  if (printed) fprintf(stdout, "%s", om_separator);
+                  if (do_colour) fprintf(stdout, "%c[%sm", 0x1b, colour_string);
+                  FWRITE(matchptr + offsets[n*2], 1, plen, stdout);
+                  if (do_colour) fprintf(stdout, "%c[00m", 0x1b);
+                  printed = TRUE;
+                  }
                 }
               }
+          
+            if (printed || printname != NULL || number) fprintf(stdout, "\n");
             }
-
-          if (printed || printname != NULL || number) fprintf(stdout, "\n");
-          }
-
-        /* Prepare to repeat to find the next match */
+          }   
+
+        /* Prepare to repeat to find the next match. If the patterned contained 
+        a lookbehind tht included \K, it is possible that the end of the match 
+        might be at or before the actual strting offset we have just used. We 
+        need to start one character further on. Unfortunately, for unanchored 
+        patterns, the actual start offset can be greater that the one that was 
+        set as a result of "bumpalong". PCRE1 does not return the actual start 
+        offset, so we have to check against the original start offset. This may 
+        lead to duplicates - we we need the fudge above to avoid printing them. 
+        (PCRE2 does this better.) */
 
         match = FALSE;
         if (line_buffered) fflush(stdout);
         rc = 0;                      /* Had some success */
         startoffset = offsets[1];    /* Restart after the match */
+        if (startoffset <= oldstartoffset)
+          {
+          if ((size_t)startoffset >= length) 
+            goto END_ONE_MATCH;              /* We were at the end */
+          startoffset = oldstartoffset + 1;
+          if (utf8)
+            while ((matchptr[startoffset] & 0xc0) == 0x80) startoffset++;    
+          }   
         goto ONLY_MATCHING_RESTART;
         }
       }
@@ -1974,6 +2008,7 @@ while (ptr < endptr)
   /* Advance to after the newline and increment the line number. The file
   offset to the current line is maintained in filepos. */
 
+  END_ONE_MATCH:
   ptr += linelength + endlinelength;
   filepos += (int)(linelength + endlinelength);
   linenumber++;
diff --git a/testdata/grepoutput b/testdata/grepoutput
index 9bf9d9d..4d61752 100644
--- a/testdata/grepoutput
+++ b/testdata/grepoutput
@@ -743,3 +743,11 @@ RC=0
 ---------------------------- Test 106 -----------------------------
 a
 RC=0
+---------------------------- Test 107 -----------------------------
+1:0,1
+2:0,1
+2:1,1
+2:2,1
+2:3,1
+2:4,1
+RC=0
-- 
2.1.0