See https://github.com/hunspell/hunspell/pull/521

Fix chenc and its usages
chenc was changing buffer it was passed to. This caused untrackable
multiple conversions of token and possibly other variables.

Change it to only return converted string

Additionally logic extending dest buffer implicitly assumed that 0
bytes are left in dest buffer. It's not necessarily the case when
converting to UTF-8 as if result would be 2-byte sequence and only 1 byte
is remaining you get E2BIG as well.

This fixes the case of pipe (-a) in UTF-8 with KOI8-R dictionary.

diff -ur hunspell-1.6.1/src/tools/hunspell.cxx hunspell-1.6.1-mod2/src/tools/hunspell.cxx
--- hunspell-1.6.1/src/tools/hunspell.cxx	2017-03-25 22:20:45.000000000 +0100
+++ hunspell-1.6.1-mod2/src/tools/hunspell.cxx	2017-08-14 23:22:16.246966174 +0200
@@ -243,7 +243,7 @@
 #endif
 
 /* change character encoding */
-std::string& chenc(std::string& st, const char* enc1, const char* enc2) {
+std::string chenc(const std::string& st, const char* enc1, const char* enc2) {
 #ifndef HAVE_ICONV
   (void)enc1;
   (void)enc2;
@@ -258,7 +258,7 @@
   std::string out(st.size(), std::string::value_type());
   size_t c1(st.size());
   size_t c2(out.size());
-  ICONV_CONST char* source = &st[0];
+  ICONV_CONST char* source = (ICONV_CONST char*) &st[0];
   char* dest = &out[0];
   iconv_t conv = iconv_open(fix_encoding_name(enc2), fix_encoding_name(enc1));
   if (conv == (iconv_t)-1) {
@@ -267,9 +267,10 @@
     size_t res;
     while ((res = iconv(conv, &source, &c1, &dest, &c2)) == size_t(-1)) {
       if (errno == E2BIG) {
+        ssize_t destoff = dest - const_cast<char*>(&out[0]);
         out.resize(out.size() + (c2 += c1));
 
-        dest = const_cast<char*>(&out[0]) + out.size() - c2;
+        dest = const_cast<char*>(&out[0]) + destoff;
       } else
         break;
     }
@@ -278,7 +279,7 @@
     }
     iconv_close(conv);
     out.resize(dest - &out[0]);
-    st = out;
+    return out;
   }
 
   return st;
@@ -507,8 +508,7 @@
 #endif
 
 int putdic(const std::string& in_word, Hunspell* pMS) {
-  std::string word(in_word);
-  chenc(word, ui_enc, dic_enc[0]);
+  std::string word = chenc(in_word, ui_enc, dic_enc[0]);
 
   std::string buf;
   pMS->input_conv(word.c_str(), buf);
@@ -565,7 +565,7 @@
   if (!dic)
     return 0;
   for (size_t i = 0; i < w.size(); ++i) {
-    chenc(w[i], io_enc, ui_enc);
+    w[i] = chenc(w[i], io_enc, ui_enc);
     fprintf(dic, "%s\n", w[i].c_str());
   }
   fclose(dic);
@@ -595,8 +595,7 @@
 // check words in the dictionaries (and set first checked dictionary)
 bool check(Hunspell** pMS, int* d, const std::string& token, int* info, std::string* root) {
   for (int i = 0; i < dmax; ++i) {
-    std::string buf(token);
-    chenc(buf, io_enc, dic_enc[*d]);
+    std::string buf = chenc(token, io_enc, dic_enc[*d]);
     mystrrep(buf, ENTITY_APOS, "'");
     if (checkapos && buf.find('\'') != std::string::npos)
       return false;
@@ -937,7 +936,7 @@
                 fprintf(stdout, "%s", chenc(wlst[0], dic_enc[d], io_enc).c_str());
               }
               for (size_t j = 1; j < wlst.size(); ++j) {
-                fprintf(stdout, ", %s", chenc(wlst[j], dic_enc[d], io_enc).c_str());
+                  fprintf(stdout, ", %s", chenc(wlst[j], dic_enc[d], io_enc).c_str());
               }
               fprintf(stdout, "\n");
               fflush(stdout);
@@ -1194,8 +1193,7 @@
 }
 
 std::string lower_first_char(const std::string& token, const char* ioenc, int langnum) {
-  std::string utf8str(token);
-  chenc(utf8str, ioenc, "UTF-8");
+  std::string utf8str = chenc(token, ioenc, "UTF-8");
   std::vector<w_char> u;
   u8_u16(u, utf8str);
   if (!u.empty()) {
@@ -1206,8 +1204,7 @@
   }
   std::string scratch;
   u16_u8(scratch, u);
-  chenc(scratch, "UTF-8", ioenc);
-  return scratch;
+  return chenc(scratch, "UTF-8", ioenc);
 }
 
 // for terminal interface
@@ -1532,13 +1529,13 @@
       std::vector<std::string> wlst;
       dialogscreen(parser, token, filename, info, wlst);  // preview
       refresh();
-      std::string buf(token);
-      wlst = pMS[d]->suggest(mystrrep(chenc(buf, io_enc, dic_enc[d]), ENTITY_APOS, "'").c_str());
+      std::string dicbuf = chenc(token, io_enc, dic_enc[d]);
+      wlst = pMS[d]->suggest(mystrrep(dicbuf, ENTITY_APOS, "'").c_str());
       if (wlst.empty()) {
         dialogexit = dialog(parser, pMS[d], token, filename, wlst, info);
       } else {
         for (size_t j = 0; j < wlst.size(); ++j) {
-          chenc(wlst[j], dic_enc[d], io_enc);
+          wlst[j] = chenc(wlst[j], dic_enc[d], io_enc);
         }
         dialogexit = dialog(parser, pMS[d], token, filename, wlst, info);
       }