Revision: 749
http://www.exim.org/viewvc/pcre2?view=rev&revision=749
Author: ph10
Date: 2017-04-14 13:39:41 +0100 (Fri, 14 Apr 2017)
Log Message:
-----------
Ensure ovector[0,1] are both PCRE2_UNSET during callouts.
Modified Paths:
--------------
code/trunk/doc/pcre2callout.3
code/trunk/src/pcre2_match.c
Modified: code/trunk/doc/pcre2callout.3
===================================================================
--- code/trunk/doc/pcre2callout.3 2017-04-14 12:14:41 UTC (rev 748)
+++ code/trunk/doc/pcre2callout.3 2017-04-14 12:39:41 UTC (rev 749)
@@ -1,4 +1,4 @@
-.TH PCRE2CALLOUT 3 "29 March 2017" "PCRE2 10.30"
+.TH PCRE2CALLOUT 3 "14 April 2017" "PCRE2 10.30"
.SH NAME
PCRE2 - Perl-compatible regular expressions (revised API)
.SH SYNOPSIS
@@ -191,8 +191,8 @@
.rs
.sp
During matching, when PCRE2 reaches a callout point, if an external function is
-set in the match context, it is called. This applies to both normal and DFA
-matching. The first argument to the callout function is a pointer to a
+provided in the match context, it is called. This applies to both normal and
+DFA matching. The first argument to the callout function is a pointer to a
\fBpcre2_callout\fP block. The second argument is the void * callout data that
was supplied when the callout was set up by calling \fBpcre2_set_callout()\fP
(see the
@@ -261,7 +261,7 @@
callout.
.P
The \fIoffset_vector\fP field is a pointer to a vector of capturing offsets
-(the "ovector"). You may read certain elements in this vector, but you must not
+(the "ovector"). You may read the elements in this vector, but you must not
change any of them.
.P
For calls to \fBpcre2_match()\fP, the \fIoffset_vector\fP field is not (since
@@ -281,10 +281,11 @@
.P
The contents of ovector[2] to ovector[<capture_top>*2-1] can be inspected in
order to extract substrings that have been matched so far, in the same way as
-extracting substrings after a match has completed. The values in ovector[0] and
-ovector[1] are undefined and should not be used in any way. Substrings that
-have not been captured (but whose numbers are less than \fIcapture_top\fP) have
-both of their ovector slots set to PCRE2_UNSET.
+extracting substrings after a match has completed. The values in ovector[0] and
+ovector[1] are always PCRE2_UNSET because the match is by definition not
+complete. Substrings that have not been captured but whose numbers are less
+than \fIcapture_top\fP also have both of their ovector slots set to
+PCRE2_UNSET.
.P
For DFA matching, the \fIoffset_vector\fP field points to the ovector that was
passed to the matching function in the match data block, but it holds no useful
@@ -410,6 +411,6 @@
.rs
.sp
.nf
-Last updated: 29 March 2017
+Last updated: 14 April 2017
Copyright (c) 1997-2017 University of Cambridge.
.fi
Modified: code/trunk/src/pcre2_match.c
===================================================================
--- code/trunk/src/pcre2_match.c 2017-04-14 12:14:41 UTC (rev 748)
+++ code/trunk/src/pcre2_match.c 2017-04-14 12:39:41 UTC (rev 749)
@@ -263,20 +263,14 @@
static int
do_callout(heapframe *F, match_block *mb, PCRE2_SIZE *lengthptr)
{
+PCRE2_SIZE save0, save1;
pcre2_callout_block cb;
+int rc;
*lengthptr = (*Fecode == OP_CALLOUT)?
PRIV(OP_lengths)[OP_CALLOUT] : GET(Fecode, 1 + 2*LINK_SIZE);
if (mb->callout == NULL) return 0; /* No callout function provided */
-
-/* The original matching code (pre 10.40) worked directly with the ovector
-passed by the user, and this was passed to callouts. Now that the working
-ovector is in the backtracking frame, it no longer needs to reserve space for
-the overall match offsets (which would waste space in the frame). For backward
-compatibility, however, we pass capture_top and offset_vector to the callout as
-if for the extended ovector. */
-
cb.version = 1;
cb.capture_top = (uint32_t)Foffset_top/2 + 1;
cb.capture_last = Fcapture_last;
@@ -304,8 +298,22 @@
cb.callout_string_length =
*lengthptr - (1 + 4*LINK_SIZE) - 2;
}
+
+/* The original matching code (pre 10.30) worked directly with the ovector
+passed by the user, and this was passed to callouts. Now that the working
+ovector is in the backtracking frame, it no longer needs to reserve space for
+the overall match offsets (which would waste space in the frame). For backward
+compatibility, however, we pass capture_top and offset_vector to the callout as
+if for the extended ovector, and we ensure that the first two slots are unset
+by preserving and restoring their current contents. */
-return mb->callout(&cb, mb->callout_data);
+save0 = Fovector[-2];
+save1 = Fovector[-1];
+Fovector[-2] = Fovector[-1] = PCRE2_UNSET;
+rc = mb->callout(&cb, mb->callout_data);
+Fovector[-2] = save0;
+Fovector[-1] = save1;
+return rc;
}
