Revision: 900
http://www.exim.org/viewvc/pcre2?view=rev&revision=900
Author: ph10
Date: 2018-01-01 14:54:06 +0000 (Mon, 01 Jan 2018)
Log Message:
-----------
Previous FIRSTLINE patch was broken. Fix it.
Modified Paths:
--------------
code/trunk/src/pcre2_dfa_match.c
code/trunk/src/pcre2_match.c
code/trunk/testdata/testinput2
code/trunk/testdata/testinput6
code/trunk/testdata/testoutput2
code/trunk/testdata/testoutput6
Modified: code/trunk/src/pcre2_dfa_match.c
===================================================================
--- code/trunk/src/pcre2_dfa_match.c 2018-01-01 14:12:35 UTC (rev 899)
+++ code/trunk/src/pcre2_dfa_match.c 2018-01-01 14:54:06 UTC (rev 900)
@@ -3363,8 +3363,6 @@
if ((re->overall_options & PCRE2_NO_START_OPTIMIZE) == 0 &&
(options & PCRE2_DFA_RESTART) == 0)
{
- PCRE2_SPTR save_end_subject = end_subject;
-
/* If firstline is TRUE, the start of the match is constrained to the first
line of a multiline string. That is, the match must be before or at the
first newline following the start of matching. Temporarily adjust
@@ -3388,13 +3386,6 @@
else
#endif
while (t < end_subject && !IS_NEWLINE(t)) t++;
-
- /* Note that we only need to advance by one code unit if we found a
- newline. If the newline is CRLF, a first code unit of LF should not
- match, because it is not at or before the newline. Similarly, only the
- first code unit of a Unicode newline might be relevant. */
-
- if (t < end_subject) t++;
end_subject = t;
}
@@ -3466,14 +3457,18 @@
#endif
}
- /* If we can't find the required code unit, break the bumpalong loop,
- to force a match failure, except when doing partial matching, when we
- let the next cycle run at the end of the subject. To see why, consider
- the pattern /(?<=abc)def/, which partially matches "abc", even though
- the string does not contain the starting character "d". */
+ /* If we can't find the required code unit, having reached the true end
+ of the subject, break the bumpalong loop, to force a match failure,
+ except when doing partial matching, when we let the next cycle run at
+ the end of the subject. To see why, consider the pattern /(?<=abc)def/,
+ which partially matches "abc", even though the string does not contain
+ the starting character "d". If we have not reached the true end of the
+ subject (PCRE2_FIRSTLINE caused end_subject to be temporarily modified)
+ we also let the cycle run, because the matching string is legitimately
+ allowed to start with the first code unit of a newline. */
if ((mb->moptions & (PCRE2_PARTIAL_HARD|PCRE2_PARTIAL_SOFT)) == 0 &&
- start_match >= end_subject)
+ start_match >= mb->end_subject)
break;
}
@@ -3532,7 +3527,7 @@
/* Restore fudged end_subject */
- end_subject = save_end_subject;
+ end_subject = mb->end_subject;
/* The following two optimizations are disabled for partial matching. */
Modified: code/trunk/src/pcre2_match.c
===================================================================
--- code/trunk/src/pcre2_match.c 2018-01-01 14:12:35 UTC (rev 899)
+++ code/trunk/src/pcre2_match.c 2018-01-01 14:54:06 UTC (rev 900)
@@ -7,7 +7,7 @@
Written by Philip Hazel
Original API code Copyright (c) 1997-2012 University of Cambridge
- New API code Copyright (c) 2015-2017 University of Cambridge
+ New API code Copyright (c) 2015-2018 University of Cambridge
-----------------------------------------------------------------------------
Redistribution and use in source and binary forms, with or without
@@ -6363,15 +6363,11 @@
if ((re->overall_options & PCRE2_NO_START_OPTIMIZE) == 0)
{
- PCRE2_SPTR save_end_subject = end_subject;
-
/* If firstline is TRUE, the start of the match is constrained to the first
line of a multiline string. That is, the match must be before or at the
first newline following the start of matching. Temporarily adjust
- end_subject so that we stop the optimization scans for a first code unit
- immediately after the first character of a newline (the first code unit can
- legitimately be a newline). If the match fails at the newline, later code
- breaks this loop. */
+ end_subject so that we stop the scans for a first code unit at a newline.
+ If the match fails at the newline, later code breaks the loop. */
if (firstline)
{
@@ -6388,13 +6384,6 @@
else
#endif
while (t < end_subject && !IS_NEWLINE(t)) t++;
-
- /* Note that we only need to advance by one code unit if we found a
- newline. If the newline is CRLF, a first code unit of LF should not
- match, because it is not at or before the newline. Similarly, only the
- first code unit of a Unicode newline might be relevant. */
-
- if (t < end_subject) t++;
end_subject = t;
}
@@ -6470,13 +6459,17 @@
#endif
}
- /* If we can't find the required code unit, break the bumpalong loop,
- to force a match failure, except when doing partial matching, when we
- let the next cycle run at the end of the subject. To see why, consider
- the pattern /(?<=abc)def/, which partially matches "abc", even though
- the string does not contain the starting character "d". */
+ /* If we can't find the required code unit, having reached the true end
+ of the subject, break the bumpalong loop, to force a match failure,
+ except when doing partial matching, when we let the next cycle run at
+ the end of the subject. To see why, consider the pattern /(?<=abc)def/,
+ which partially matches "abc", even though the string does not contain
+ the starting character "d". If we have not reached the true end of the
+ subject (PCRE2_FIRSTLINE caused end_subject to be temporarily modified)
+ we also let the cycle run, because the matching string is legitimately
+ allowed to start with the first code unit of a newline. */
- if (!mb->partial && start_match >= end_subject)
+ if (!mb->partial && start_match >= mb->end_subject)
{
rc = MATCH_NOMATCH;
break;
@@ -6538,7 +6531,7 @@
/* Restore fudged end_subject */
- end_subject = save_end_subject;
+ end_subject = mb->end_subject;
/* The following two optimizations must be disabled for partial matching. */
Modified: code/trunk/testdata/testinput2
===================================================================
--- code/trunk/testdata/testinput2 2018-01-01 14:12:35 UTC (rev 899)
+++ code/trunk/testdata/testinput2 2018-01-01 14:54:06 UTC (rev 900)
@@ -5405,4 +5405,8 @@
\= Expect no match
xyz\r\nabc
+/[abc]/firstline
+\= Expect no match
+ \na
+
# End of testinput2
Modified: code/trunk/testdata/testinput6
===================================================================
--- code/trunk/testdata/testinput6 2018-01-01 14:12:35 UTC (rev 899)
+++ code/trunk/testdata/testinput6 2018-01-01 14:54:06 UTC (rev 900)
@@ -4942,4 +4942,8 @@
\= Expect no match
xyz\r\nabc
+/[abc]/firstline
+\= Expect no match
+ \na
+
# End of testinput6
Modified: code/trunk/testdata/testoutput2
===================================================================
--- code/trunk/testdata/testoutput2 2018-01-01 14:12:35 UTC (rev 899)
+++ code/trunk/testdata/testoutput2 2018-01-01 14:54:06 UTC (rev 900)
@@ -16453,6 +16453,11 @@
xyz\r\nabc
No match
+/[abc]/firstline
+\= Expect no match
+ \na
+No match
+
# End of testinput2
Error -65: PCRE2_ERROR_BADDATA (unknown error number)
Error -62: bad serialized data
Modified: code/trunk/testdata/testoutput6
===================================================================
--- code/trunk/testdata/testoutput6 2018-01-01 14:12:35 UTC (rev 899)
+++ code/trunk/testdata/testoutput6 2018-01-01 14:54:06 UTC (rev 900)
@@ -7766,4 +7766,9 @@
xyz\r\nabc
No match
+/[abc]/firstline
+\= Expect no match
+ \na
+No match
+
# End of testinput6
