Fix incorrect computation of group length when one branch exceeded 65535.

ChangeLog

@@ -124,6 +124,11 @@ available directly).
 
 26. Make pcre2test -C show if libreadline or libedit is supported.
 
+28. If the length of one branch of a group exceeded 65535 (the maximum value
+that is remembered as a minimum length), the whole group's length was 
+incorrectly recorded as 65535, leading to incorrect "no match" when start-up 
+optimizations were in force.
+
 
 Version 10.33 16-April-2019
 ---------------------------

src/pcre2_study.c

@@ -105,6 +105,7 @@ find_minlength(const pcre2_real_code *re, PCRE2_SPTR code,
   int *backref_cache)
 {
 int length = -1;
+int branchlength = 0;
 int prev_cap_recno = -1;
 int prev_cap_d = 0;
 int prev_recurse_recno = -1;
@@ -112,9 +113,9 @@ int prev_recurse_d = 0;
 uint32_t once_fudge = 0;
 BOOL had_recurse = FALSE;
 BOOL dupcapused = (re->flags & PCRE2_DUPCAPUSED) != 0;
-recurse_check this_recurse;
-int branchlength = 0;
+PCRE2_SPTR nextbranch = code + GET(code, 1);
 PCRE2_UCHAR *cc = (PCRE2_UCHAR *)code + 1 + LINK_SIZE;
+recurse_check this_recurse;
 
 /* If this is a "could be empty" group, its minimum length is 0. */
 
@@ -130,16 +131,20 @@ if ((*countptr)++ > 1000) return -1;
 
 /* Scan along the opcodes for this branch. If we get to the end of the branch,
 check the length against that of the other branches. If the accumulated length
-passes 16-bits, stop. */
+passes 16-bits, reset to that value and skip the rest of the branch. */
 
 for (;;)
   {
   int d, min, recno;
-  PCRE2_UCHAR *cs, *ce;
-  PCRE2_UCHAR op = *cc;
+  PCRE2_UCHAR op, *cs, *ce;
 
-  if (branchlength >= UINT16_MAX) return UINT16_MAX;
+  if (branchlength >= UINT16_MAX)
+    {
+    branchlength = UINT16_MAX;
+    cc = (PCRE2_UCHAR *)nextbranch;
+    }
 
+  op = *cc;
   switch (op)
     {
     case OP_COND:
@@ -229,6 +234,7 @@ for (;;)
     if (length < 0 || (!had_recurse && branchlength < length))
       length = branchlength;
     if (op != OP_ALT) return length;
+    nextbranch = cc + GET(cc, 1);
     cc += 1 + LINK_SIZE;
     branchlength = 0;
     had_recurse = FALSE;

testdata/testinput2

@@ -5740,4 +5740,12 @@ a)"xI
 /c*+/
     ab\=ph,offset=2
 
+/\A\s*(a|(?:[^`]{28500}){4})/I
+    a
+
+/\A\s*((?:[^`]{28500}){4})/I
+
+/\A\s*((?:[^`]{28500}){4}|a)/I
+    a
+
 # End of testinput2

testdata/testoutput2

@@ -17267,6 +17267,33 @@ Partial match:
     ab\=ph,offset=2
 Partial match: 
 
+/\A\s*(a|(?:[^`]{28500}){4})/I
+Capture group count = 1
+Max lookbehind = 1
+Compile options: <none>
+Overall options: anchored
+Subject length lower bound = 1
+    a
+ 0: a
+ 1: a
+
+/\A\s*((?:[^`]{28500}){4})/I
+Capture group count = 1
+Max lookbehind = 1
+Compile options: <none>
+Overall options: anchored
+Subject length lower bound = 65535
+
+/\A\s*((?:[^`]{28500}){4}|a)/I
+Capture group count = 1
+Max lookbehind = 1
+Compile options: <none>
+Overall options: anchored
+Subject length lower bound = 1
+    a
+ 0: a
+ 1: a
+
 # End of testinput2
 Error -70: PCRE2_ERROR_BADDATA (unknown error number)
 Error -62: bad serialized data