erl_interface: Do not generate atoms on old latin1 ext format

Solved by letting ei_encode_atom_as ignore 'to_enc' argument
and always encode in UTF8 format.

4 files changed, 26 insertions, 75 deletions

diff --git a/lib/erl_interface/src/encode/encode_atom.c b/lib/erl_interface/src/encode/encode_atom.c
index c1817628e5..1fd7811a0e 100644
--- a/lib/erl_interface/src/encode/encode_atom.c
+++ b/lib/erl_interface/src/encode/encode_atom.c
@@ -26,7 +26,6 @@
 
 static int verify_ascii_atom(const char* src, int slen);
 static int verify_utf8_atom(const char* src, int slen);
-static int is_latin1_as_utf8(const char *p, int len);
 
 int ei_encode_atom(char *buf, int *index, const char *p)
 {
@@ -34,7 +33,7 @@ int ei_encode_atom(char *buf, int *index, const char *p)
 
     if (len >= MAXATOMLEN)
 	len = MAXATOMLEN - 1;
-    return ei_encode_atom_len_as(buf, index, p, len, ERLANG_LATIN1, ERLANG_LATIN1);
+    return ei_encode_atom_len_as(buf, index, p, len, ERLANG_LATIN1, 0);
 }
 
 int ei_encode_atom_len(char *buf, int *index, const char *p, int len)
@@ -42,7 +41,7 @@ int ei_encode_atom_len(char *buf, int *index, const char *p, int len)
     /* This function is documented to truncate at MAXATOMLEN (256) */ 
     if (len >= MAXATOMLEN)
 	len = MAXATOMLEN - 1;
-    return ei_encode_atom_len_as(buf, index, p, len, ERLANG_LATIN1, ERLANG_LATIN1);
+    return ei_encode_atom_len_as(buf, index, p, len, ERLANG_LATIN1, 0);
 }
 
 int ei_encode_atom_as(char *buf, int *index, const char *p,
@@ -64,46 +63,11 @@ int ei_encode_atom_len_as(char *buf, int *index, const char *p, int len,
       return -1;
   }
 
-  if (to_enc == (ERLANG_LATIN1 | ERLANG_UTF8)) {
-    if (from_enc == ERLANG_UTF8) {
-      to_enc = is_latin1_as_utf8(p, len) ? ERLANG_LATIN1 : ERLANG_UTF8;
-    }
-    else {
-      to_enc = from_enc;
-    }
-  }
-  switch(to_enc) {
-  case ERLANG_LATIN1:
-      if (buf) {
-	  put8(s,ERL_ATOM_EXT);
-	  switch (from_enc) {
-	  case ERLANG_UTF8:
-	      len = utf8_to_latin1(s+2, p, len, MAXATOMLEN-1, NULL);
-	      if (len < 0) return -1;
-	      break;
-	  case ERLANG_ASCII:
-	      if (verify_ascii_atom(p, len) < 0) return -1;
-	      memcpy(s+2, p, len);
-	      break;
-	  case ERLANG_LATIN1:
-	      memcpy(s+2, p, len);
-	      break;
-	  default:
-	      return -1;
-	  }
-	  put16be(s,len);
-      }
-      else {
-	  s += 3;
-	  if (from_enc == ERLANG_UTF8) {
-	      len = utf8_to_latin1(NULL, p, len, MAXATOMLEN-1, NULL);
-	      if (len < 0) return -1;
-	  } else if (from_enc == ERLANG_ASCII)
-	    if (verify_ascii_atom(p, len) < 0) return -1;
-      }
-      break;
-      
-  case ERLANG_UTF8:
+  /*
+   * Since OTP 20 we totally ignore 'to_enc'
+   * and alway encode as UTF8.
+   */
+  {
       offs =  1 + 1;
       switch (from_enc) {
       case ERLANG_LATIN1:
@@ -133,10 +97,6 @@ int ei_encode_atom_len_as(char *buf, int *index, const char *p, int len,
 	  }
       }
       else s+= offs;
-      break;
-
-  default:
-      return -1;
   }
   s += len;
 
@@ -197,13 +157,3 @@ static int verify_utf8_atom(const char* src, int slen)
     return 0;
 }
 
-/* Only latin1 code points in utf8 string?
- */
-static int is_latin1_as_utf8(const char *p, int len)
-{
-  int i;
-  for (i=0; i<len; i++) {
-    if ((unsigned char)p[i] > 0xC3) return 0;
-  }
-  return 1;
-}
diff --git a/lib/erl_interface/src/encode/encode_boolean.c b/lib/erl_interface/src/encode/encode_boolean.c
index 61e7e5e6e7..053029af05 100644
--- a/lib/erl_interface/src/encode/encode_boolean.c
+++ b/lib/erl_interface/src/encode/encode_boolean.c
@@ -32,12 +32,12 @@ int ei_encode_boolean(char *buf, int *index, int p)
   val = p ? "true" : "false";
   len = strlen(val);
 
-  if (!buf) s += 3;
+  if (!buf) s += 2;
   else {
-    put8(s,ERL_ATOM_EXT);
-    put16be(s,len);
+      put8(s, ERL_SMALL_ATOM_UTF8_EXT);
+      put8(s, len);
 
-    memmove(s,val,len); /* unterminated string */
+      memcpy(s,val,len); /* unterminated string */
   }
   s += len;
 
diff --git a/lib/erl_interface/src/legacy/erl_eterm.c b/lib/erl_interface/src/legacy/erl_eterm.c
index e4b3b49c7d..5153d0f2e7 100644
--- a/lib/erl_interface/src/legacy/erl_eterm.c
+++ b/lib/erl_interface/src/legacy/erl_eterm.c
@@ -188,14 +188,20 @@ char* erl_atom_ptr_latin1(Erl_Atom_data* a)
 char* erl_atom_ptr_utf8(Erl_Atom_data* a)
 {
     if (a->utf8 == NULL) {
-	int dlen = a->lenL * 2; /* over estimation */
-	a->utf8 = malloc(dlen + 1); 
-	a->lenU = latin1_to_utf8(a->utf8, a->latin1, a->lenL, dlen, NULL);
-	a->utf8[a->lenU] = '\0';
+        erlang_char_encoding enc;
+        a->lenU = latin1_to_utf8(NULL, a->latin1, a->lenL, a->lenL*2, &enc);
+        if (enc == ERLANG_ASCII) {
+            a->utf8 = a->latin1;
+        }
+        else {
+            a->utf8 = malloc(a->lenU + 1);
+            latin1_to_utf8(a->utf8, a->latin1, a->lenL, a->lenU, NULL);
+            a->utf8[a->lenU] = '\0';
+        }
     }
     return a->utf8;
-
 }
+
 int erl_atom_size_latin1(Erl_Atom_data* a)
 {
     if (a->latin1 == NULL) {
diff --git a/lib/erl_interface/src/legacy/erl_marshal.c b/lib/erl_interface/src/legacy/erl_marshal.c
index 527ae0ef8f..b7a8455313 100644
--- a/lib/erl_interface/src/legacy/erl_marshal.c
+++ b/lib/erl_interface/src/legacy/erl_marshal.c
@@ -175,10 +175,9 @@ static void encode_atom(Erl_Atom_data* a, unsigned char **ext)
     int ix = 0;
     if (a->latin1) {
 	ei_encode_atom_len_as((char*)*ext, &ix, a->latin1, a->lenL,
-			      ERLANG_LATIN1, ERLANG_LATIN1);
+			      ERLANG_LATIN1, ERLANG_UTF8);
     }
-    else if (ei_encode_atom_len_as((char*)*ext, &ix, a->utf8, a->lenU,
-				   ERLANG_UTF8, ERLANG_LATIN1) < 0) {
+    else {
 	ei_encode_atom_len_as((char*)*ext, &ix, a->utf8, a->lenU,
 			      ERLANG_UTF8, ERLANG_UTF8);
     }
@@ -542,12 +541,8 @@ int erl_term_len(ETERM *ep)
 
 static int atom_len_helper(Erl_Atom_data* a)
 {
-    if (erl_atom_ptr_latin1(a)) {
-	return 1 + 2 + a->lenL; /* ERL_ATOM_EXT */
-    }
-    else {
-	return 1 + 1 + (a->lenU > 255) + a->lenU;
-    }
+    (void) erl_atom_ptr_utf8(a);
+    return 1 + 1 + (a->lenU > 255) + a->lenU;
 }
 
 static int erl_term_len_helper(ETERM *ep, int dist)