vim-patch:7.4.2223

Problem:    Buffer overflow when using latin1 character with feedkeys().
Solution:   Check for an illegal character.  Add a test.

d3c907b5d2

src/nvim/eval.c

@@ -16932,7 +16932,7 @@ static void f_strgetchar(typval_T *argvars, typval_T *rettv, FunPtr fptr)
         break;
       }
       charidx--;
-      byteidx += mb_cptr2len(str + byteidx);
+      byteidx += MB_CPTR2LEN(str + byteidx);
     }
   }
 }
@@ -17054,7 +17054,7 @@ static void f_strcharpart(typval_T *argvars, typval_T *rettv, FunPtr fptr) {
   if (!error) {
     if (nchar > 0) {
       while (nchar > 0 && nbyte < slen) {
-        nbyte += mb_cptr2len(p + nbyte);
+        nbyte += MB_CPTR2LEN(p + nbyte);
         nchar--;
       }
     } else {
@@ -17069,7 +17069,7 @@ static void f_strcharpart(typval_T *argvars, typval_T *rettv, FunPtr fptr) {
       if (off < 0) {
         len += 1;
       } else {
-        len += mb_cptr2len(p + off);
+        len += MB_CPTR2LEN(p + off);
       }
       charlen--;
     }

src/nvim/getchar.c

@@ -3746,8 +3746,10 @@ eval_map_expr (
  */
 char_u *vim_strsave_escape_csi(char_u *p)
 {
-  /* Need a buffer to hold up to three times as much. */
-  char_u *res = xmalloc(STRLEN(p) * 3 + 1);
+  // Need a buffer to hold up to three times as much.  Four in case of an
+  // illegal utf-8 byte:
+  // 0xc0 -> 0xc3 - 0x80 -> 0xc3 K_SPECIAL KS_SPECIAL KE_FILLER
+  char_u *res = xmalloc(STRLEN(p) * 4 + 1);
   char_u *d = res;
   for (char_u *s = p; *s != NUL; ) {
     if (s[0] == K_SPECIAL && s[1] != NUL && s[2] != NUL) {
@@ -3756,17 +3758,10 @@ char_u *vim_strsave_escape_csi(char_u *p)
       *d++ = *s++;
       *d++ = *s++;
     } else {
-      int len  = mb_char2len(PTR2CHAR(s));
-      int len2 = mb_ptr2len(s);
-      /* Add character, possibly multi-byte to destination, escaping
-       * CSI and K_SPECIAL. */
+      // Add character, possibly multi-byte to destination, escaping
+      // CSI and K_SPECIAL. Be careful, it can be an illegal byte!
       d = add_char2buf(PTR2CHAR(s), d);
-      while (len < len2) {
-        /* add following combining char */
-        d = add_char2buf(PTR2CHAR(s + len), d);
-        len += mb_char2len(PTR2CHAR(s + len));
-      }
-      mb_ptr_adv(s);
+      s += MB_CPTR2LEN(s);
     }
   }
   *d = NUL;

src/nvim/macros.h

@@ -138,7 +138,7 @@
 // Backup multi-byte pointer. Only use with "p" > "s" !
 # define mb_ptr_back(s, p)  (p -= mb_head_off((char_u *)s, (char_u *)p - 1) + 1)
 // get length of multi-byte char, not including composing chars
-# define mb_cptr2len(p)     utf_ptr2len(p)
+# define MB_CPTR2LEN(p)     utf_ptr2len(p)
 
 # define MB_COPY_CHAR(f, t) mb_copy_char((const char_u **)(&f), &t);
 

src/nvim/spell.c

@@ -4507,7 +4507,7 @@ static void suggest_trie_walk(suginfo_T *su, langp_T *lp, char_u *fword, bool so
       }
 
       if (has_mbyte) {
-        n = mb_cptr2len(p);
+        n = MB_CPTR2LEN(p);
         c = mb_ptr2char(p);
         if (p[n] == NUL)
           c2 = NUL;
@@ -4584,9 +4584,9 @@ static void suggest_trie_walk(suginfo_T *su, langp_T *lp, char_u *fword, bool so
       // "fword" here, it's changed back afterwards at STATE_UNSWAP3.
       p = fword + sp->ts_fidx;
       if (has_mbyte) {
-        n = mb_cptr2len(p);
+        n = MB_CPTR2LEN(p);
         c = mb_ptr2char(p);
-        fl = mb_cptr2len(p + n);
+        fl = MB_CPTR2LEN(p + n);
         c2 = mb_ptr2char(p + n);
         if (!soundfold && !spell_iswordp(p + n + fl, curwin))
           c3 = c;               // don't swap non-word char
@@ -4682,10 +4682,10 @@ static void suggest_trie_walk(suginfo_T *su, langp_T *lp, char_u *fword, bool so
         ++depth;
         p = fword + sp->ts_fidx;
         if (has_mbyte) {
-          n = mb_cptr2len(p);
+          n = MB_CPTR2LEN(p);
           c = mb_ptr2char(p);
-          fl = mb_cptr2len(p + n);
-          fl += mb_cptr2len(p + n + fl);
+          fl = MB_CPTR2LEN(p + n);
+          fl += MB_CPTR2LEN(p + n + fl);
           memmove(p, p + n, fl);
           mb_char2bytes(c, p + fl);
           stack[depth].ts_fidxtry = sp->ts_fidx + n + fl;
@@ -4734,10 +4734,10 @@ static void suggest_trie_walk(suginfo_T *su, langp_T *lp, char_u *fword, bool so
         ++depth;
         p = fword + sp->ts_fidx;
         if (has_mbyte) {
-          n = mb_cptr2len(p);
-          n += mb_cptr2len(p + n);
+          n = MB_CPTR2LEN(p);
+          n += MB_CPTR2LEN(p + n);
           c = mb_ptr2char(p + n);
-          tl = mb_cptr2len(p + n);
+          tl = MB_CPTR2LEN(p + n);
           memmove(p + tl, p, n);
           mb_char2bytes(c, p);
           stack[depth].ts_fidxtry = sp->ts_fidx + n + tl;
@@ -4980,8 +4980,8 @@ static void find_keepcap_word(slang_T *slang, char_u *fword, char_u *kword)
       // round[depth] == 1: Try using the folded-case character.
       // round[depth] == 2: Try using the upper-case character.
       if (has_mbyte) {
-        flen = mb_cptr2len(fword + fwordidx[depth]);
-        ulen = mb_cptr2len(uword + uwordidx[depth]);
+        flen = MB_CPTR2LEN(fword + fwordidx[depth]);
+        ulen = MB_CPTR2LEN(uword + uwordidx[depth]);
       } else
         ulen = flen = 1;
       if (round[depth] == 1) {

src/nvim/testdir/test_alot.vim

@@ -21,6 +21,7 @@ source test_options.vim
 source test_partial.vim
 source test_popup.vim
 source test_regexp_utf8.vim
+source test_source_utf8.vim
 source test_statusline.vim
 source test_syn_attr.vim
 source test_tabline.vim

src/nvim/testdir/test_regexp_utf8.vim

@@ -90,21 +90,6 @@ func Test_classes_re2()
   set re=0
 endfunc
 
-func Test_source_utf8()
-  " check that sourcing a script with 0x80 as second byte works
-  new
-  call setline(1, [':%s/àx/--à1234--/g', ':%s/Àx/--À1234--/g'])
-  write! Xscript
-  bwipe!
-  new
-  call setline(1, [' àx ', ' Àx '])
-  source! Xscript | echo
-  call assert_equal(' --à1234-- ', getline(1))
-  call assert_equal(' --À1234-- ', getline(2))
-  bwipe!
-  call delete('Xscript')
-endfunc
-
 func Test_recursive_substitute()
   new
   s/^/\=execute("s#^##gn")

src/nvim/testdir/test_source_utf8.vim

@@ -0,0 +1,33 @@
+" Test the :source! command
+if !has('multi_byte')
+  finish
+endif
+
+func Test_source_utf8()
+  " check that sourcing a script with 0x80 as second byte works
+  new
+  call setline(1, [':%s/àx/--à1234--/g', ':%s/Àx/--À1234--/g'])
+  write! Xscript
+  bwipe!
+  new
+  call setline(1, [' àx ', ' Àx '])
+  source! Xscript | echo
+  call assert_equal(' --à1234-- ', getline(1))
+  call assert_equal(' --À1234-- ', getline(2))
+  bwipe!
+  call delete('Xscript')
+endfunc
+
+func Test_source_latin()
+  " check that sourcing a latin1 script with a 0xc0 byte works
+  new
+  call setline(1, ["call feedkeys('r')", "call feedkeys('\xc0', 'xt')"])
+  write! Xscript
+  bwipe!
+  new
+  call setline(1, ['xxx'])
+  source Xscript
+  call assert_equal("\u00c0xx", getline(1))
+  bwipe!
+  call delete('Xscript')
+endfunc

src/nvim/version.c

@@ -217,7 +217,7 @@ static int included_patches[] = {
   2226,
   2225,
   // 2224,
-  // 2223,
+  2223,
   2222,
   // 2221,
   2220,