UTF-8 support for distribution

1 files changed, 120 insertions, 73 deletions

diff --git a/erts/emulator/beam/external.c b/erts/emulator/beam/external.c
index 68edcd0fa6..8c4d9108d4 100644
--- a/erts/emulator/beam/external.c
+++ b/erts/emulator/beam/external.c
@@ -142,6 +142,7 @@ erts_init_atom_cache_map(ErtsAtomCacheMap *acmp)
 {
     if (acmp) {
 	int ix;
+	acmp->long_atoms = 0;
 	for (ix = 0; ix < ERTS_ATOM_CACHE_SIZE; ix++)
 	    acmp->cache[ix].iix = -1;
 	acmp->sz = 0;
@@ -154,6 +155,7 @@ erts_reset_atom_cache_map(ErtsAtomCacheMap *acmp)
 {
     if (acmp) {
 	int i;
+	acmp->long_atoms = 0;
 	for (i = 0; i < acmp->sz; i++) {
 	    ASSERT(0 <= acmp->cix[i] && acmp->cix[i] < ERTS_ATOM_CACHE_SIZE);
 	    acmp->cache[acmp->cix[i]].iix = -1;
@@ -175,9 +177,23 @@ erts_destroy_atom_cache_map(ErtsAtomCacheMap *acmp)
 }
 
 static ERTS_INLINE void
-insert_acache_map(ErtsAtomCacheMap *acmp, Eterm atom)
+insert_acache_map(ErtsAtomCacheMap *acmp, Eterm atom, Uint32 dflags)
 {
-    if (acmp && acmp->sz < ERTS_MAX_INTERNAL_ATOM_CACHE_ENTRIES) {
+    /*
+     * If the receiver do not understand utf8 atoms
+     * and this atom cannot be represented in latin1,
+     * we are not allowed to cache it.
+     *
+     * In this case all atoms are assumed to have
+     * latin1 encoding in the cache. By refusing it
+     * in the cache we will instead encode it using
+     * ATOM_UTF8_EXT/SMALL_ATOM_UTF8_EXT which the
+     * receiver do not recognize and tear down the
+     * connection.
+     */
+    if (acmp && acmp->sz < ERTS_MAX_INTERNAL_ATOM_CACHE_ENTRIES
+	&& ((dflags & DFLAG_UTF8_ATOMS)
+	    || atom_tab(atom_val(atom))->latin1_chars >= 0)) {
 	int ix;
 	ASSERT(acmp->hdr_sz < 0);
 	ix = atom2cix(atom);
@@ -190,7 +206,7 @@ insert_acache_map(ErtsAtomCacheMap *acmp, Eterm atom)
 }
 
 static ERTS_INLINE int
-get_iix_acache_map(ErtsAtomCacheMap *acmp, Eterm atom)
+get_iix_acache_map(ErtsAtomCacheMap *acmp, Eterm atom, Uint32 dflags)
 {
     if (!acmp)
 	return -1;
@@ -199,7 +215,9 @@ get_iix_acache_map(ErtsAtomCacheMap *acmp, Eterm atom)
 	ASSERT(is_atom(atom));
 	ix = atom2cix(atom);
 	if (acmp->cache[ix].iix < 0) {
-	    ASSERT(acmp->sz == ERTS_MAX_INTERNAL_ATOM_CACHE_ENTRIES);
+	    ASSERT(acmp->sz == ERTS_MAX_INTERNAL_ATOM_CACHE_ENTRIES
+		   || (!(dflags & DFLAG_UTF8_ATOMS)
+		       && atom_tab(atom_val(atom))->latin1_chars < 0));
 	    return -1;
 	}
 	else {
@@ -210,18 +228,17 @@ get_iix_acache_map(ErtsAtomCacheMap *acmp, Eterm atom)
 }
 
 void
-erts_finalize_atom_cache_map(ErtsAtomCacheMap *acmp)
+erts_finalize_atom_cache_map(ErtsAtomCacheMap *acmp, Uint32 dflags)
 {
     if (acmp) {
-#if MAX_ATOM_LENGTH > 255
-#error "This code is not complete; long_atoms info need to be passed to the following stages."
-	int long_atoms = 0; /* !0 if one or more atoms are long than 255. */
-#endif
+	int utf8_atoms = (int) (dflags & DFLAG_UTF8_ATOMS);
+	int long_atoms = 0; /* !0 if one or more atoms are longer than 255. */
 	int i;
 	int sz;
 	int fix_sz
 	    = 1 /* VERSION_MAGIC */
 	    + 1 /* DIST_HEADER */
+	    + 1 /* dist header flags */
 	    + 1 /* number of internal cache entries */
 	    ;
 	int min_sz;
@@ -230,22 +247,23 @@ erts_finalize_atom_cache_map(ErtsAtomCacheMap *acmp)
 	min_sz = fix_sz+(2+4)*acmp->sz;
 	sz = fix_sz;
 	for (i = 0; i < acmp->sz; i++) {
+	    Atom *a;
 	    Eterm atom;
 	    int len;
 	    atom = acmp->cache[acmp->cix[i]].atom;
 	    ASSERT(is_atom(atom));
-	    len = atom_tab(atom_val(atom))->len;
-#if MAX_ATOM_LENGTH > 255
+	    a = atom_tab(atom_val(atom));
+	    len = (int) (utf8_atoms ? a->len : a->latin1_chars);
+	    ASSERT(len >= 0);
 	    if (!long_atoms && len > 255)
 		long_atoms = 1;
-#endif
 	    /* Enough for a new atom cache value */
 	    sz += 1 /* cix */ + 1 /* length */ + len /* text */;
 	}
-#if MAX_ATOM_LENGTH > 255
-	if (long_atoms)
+	if (long_atoms) {
+	    acmp->long_atoms = 1;
 	    sz += acmp->sz; /* we need 2 bytes per atom for length */
-#endif
+	}
 	/* Dynamically sized flag field */
 	sz += ERTS_DIST_HDR_ATOM_CACHE_FLAG_BYTES(acmp->sz);
 	if (sz < min_sz)
@@ -274,6 +292,7 @@ byte *erts_encode_ext_dist_header_setup(byte *ctl_ext, ErtsAtomCacheMap *acmp)
     else {
 	int i;
 	byte *ep = ctl_ext;
+	byte dist_hdr_flags = acmp->long_atoms ? ERTS_DIST_HDR_LONG_ATOMS_FLG : 0;
 	ASSERT(acmp->hdr_sz >= 0);
 	/*
 	 * Write cache update instructions. Note that this is a purely
@@ -296,28 +315,36 @@ byte *erts_encode_ext_dist_header_setup(byte *ctl_ext, ErtsAtomCacheMap *acmp)
 	}
 	--ep;
 	put_int8(acmp->sz, ep);
+	--ep;
+	put_int8(dist_hdr_flags, ep);
 	*--ep = DIST_HEADER;
 	*--ep = VERSION_MAGIC;
 	return ep;
     }
 }
 
-byte *erts_encode_ext_dist_header_finalize(byte *ext, ErtsAtomCache *cache)
+byte *erts_encode_ext_dist_header_finalize(byte *ext, ErtsAtomCache *cache, Uint32 dflags)
 {
     byte *ip;
     byte instr_buf[(2+4)*ERTS_ATOM_CACHE_SIZE];
     int ci, sz;
+    byte dist_hdr_flags;
+    int long_atoms;
+    int utf8_atoms = (int) (dflags & DFLAG_UTF8_ATOMS);
     register byte *ep = ext;
     ASSERT(ep[0] == VERSION_MAGIC);
     if (ep[1] != DIST_HEADER)
 	return ext;
 
+    dist_hdr_flags = ep[2];
+    long_atoms = ERTS_DIST_HDR_LONG_ATOMS_FLG & ((int) dist_hdr_flags);
+
     /*
      * Update output atom cache and write the external version of
      * the dist header. We write the header backwards just
      * before the actual term(s).
      */
-    ep += 2;
+    ep += 3;
     ci = (int) get_int8(ep);
     ASSERT(0 <= ci && ci < ERTS_ATOM_CACHE_SIZE);
     ep += 1;
@@ -342,12 +369,7 @@ byte *erts_encode_ext_dist_header_finalize(byte *ext, ErtsAtomCache *cache)
 	flgs_bytes = ERTS_DIST_HDR_ATOM_CACHE_FLAG_BYTES(ci);
 
 	ASSERT(flgs_bytes <= sizeof(flgs_buf));
-#if MAX_ATOM_LENGTH > 255
-	/* long_atoms info needs to be passed from previous stages */
-	if (long_atoms)
-	    flgs |= ERTS_DIST_HDR_LONG_ATOMS_FLG;
-#endif
-	flgs = 0;
+	flgs = (Uint32) dist_hdr_flags;
 	flgs_buf_ix = 0;
 	if ((ci & 1) == 0)
 	    used_half_bytes = 2;
@@ -382,17 +404,22 @@ byte *erts_encode_ext_dist_header_finalize(byte *ext, ErtsAtomCache *cache)
 		Atom *a;
 		cache->out_arr[cix] = atom;
 		a = atom_tab(atom_val(atom));
-		sz = a->len;
-		ep -= sz;
-		sys_memcpy((void *) ep, (void *) a->name, sz);
-#if MAX_ATOM_LENGTH > 255
+		if (utf8_atoms) {
+		    sz = a->len;
+		    ep -= sz;
+		    sys_memcpy((void *) ep, (void *) a->name, sz);
+		}
+		else {
+		    ASSERT(0 <= a->latin1_chars && a->latin1_chars <= MAX_ATOM_CHARACTERS);
+		    ep -= a->latin1_chars;
+		    sz = erts_utf8_to_latin1(ep, a->name, a->len);
+		    ASSERT(a->latin1_chars == sz);
+		}
 		if (long_atoms) {
 		    ep -= 2;
 		    put_int16(sz, ep);
 		}
-		else
-#endif
-		{
+		else {
 		    ASSERT(0 <= sz && sz <= 255);
 		    --ep;
 		    put_int8(sz, ep);
@@ -553,6 +580,7 @@ erts_prepare_dist_ext(ErtsDistExternal *edep,
 #endif
 
     register byte *ep = ext;
+    int utf8_atoms = (int) (dep->flags & DFLAG_UTF8_ATOMS);
 
     edep->heap_size = -1;
     edep->ext_endp = ext+size;
@@ -611,9 +639,7 @@ erts_prepare_dist_ext(ErtsDistExternal *edep,
 	    ERTS_EXT_HDR_FAIL;
 	ep++;
 	if (no_atoms) {
-#if MAX_ATOM_LENGTH > 255
 	    int long_atoms = 0;
-#endif
 #ifdef DEBUG
 	    byte *flgs_buf = ep;
 #endif
@@ -632,14 +658,8 @@ erts_prepare_dist_ext(ErtsDistExternal *edep,
 	     */
 	    byte_ix = ERTS_DIST_HDR_ATOM_CACHE_FLAG_BYTE_IX(no_atoms);
 	    bit_ix = ERTS_DIST_HDR_ATOM_CACHE_FLAG_BIT_IX(no_atoms);
-	    if (flgsp[byte_ix] & (((byte) ERTS_DIST_HDR_LONG_ATOMS_FLG)
-				  << bit_ix)) {
-#if MAX_ATOM_LENGTH > 255
+	    if (flgsp[byte_ix] & (((byte) ERTS_DIST_HDR_LONG_ATOMS_FLG) << bit_ix))
 		long_atoms = 1;
-#else
-		ERTS_EXT_HDR_FAIL; /* Long atoms not supported yet */
-#endif
-	    }
 
 #ifdef DEBUG
 	    byte_ix = 0;
@@ -707,23 +727,25 @@ erts_prepare_dist_ext(ErtsDistExternal *edep,
 		    if (cix >= ERTS_ATOM_CACHE_SIZE)
 			ERTS_EXT_HDR_FAIL;
 		    ep++;
-#if MAX_ATOM_CHARACTERS > 255
 		    if (long_atoms) {
 			CHKSIZE(2);
 			len = get_int16(ep);
 			ep += 2;
 		    }
-		    else
-#endif
-		    {
+		    else {
 			CHKSIZE(1);
 			len = get_int8(ep);
 			ep++;
 		    }
-		    if (len > MAX_ATOM_CHARACTERS)
-			ERTS_EXT_HDR_FAIL; /* Too long atom */
 		    CHKSIZE(len);
-		    atom = am_atom_put((char *) ep, len);
+		    atom = erts_atom_put((byte *) ep,
+					 len,
+					 (utf8_atoms
+					  ? ERTS_ATOM_ENC_UTF8
+					  : ERTS_ATOM_ENC_LATIN1),
+					 0);
+		    if (is_non_value(atom))
+			ERTS_EXT_HDR_FAIL;
 		    ep += len;
 		    cache->in_arr[cix] = atom;
 		    edep->attab.atom[tix] = atom;
@@ -1404,7 +1426,8 @@ static byte*
 enc_atom(ErtsAtomCacheMap *acmp, Eterm atom, byte *ep, Uint32 dflags)
 {
     int iix;
-    int i, j;
+    int len;
+    int utf8_atoms = (int) (dflags & DFLAG_UTF8_ATOMS);
 
     ASSERT(is_atom(atom));
 
@@ -1423,42 +1446,46 @@ enc_atom(ErtsAtomCacheMap *acmp, Eterm atom, byte *ep, Uint32 dflags)
 	}
 	return ep;
     }
+
     /*
      * term_to_binary/1,2 and the initial distribution message
      * don't use the cache.
      */
-    iix = get_iix_acache_map(acmp, atom);
-    if (iix < 0) { 
-	i = atom_val(atom);
-	j = atom_tab(i)->len;
-	if (dflags & DFLAG_UTF8_ATOMS) {
-	    if (j <= 255) {
+
+    iix = get_iix_acache_map(acmp, atom, dflags);
+    if (iix < 0) {
+	Atom *a = atom_tab(atom_val(atom));
+	if (utf8_atoms || a->latin1_chars < 0) {
+	    len = a->len;
+	    if (len > 255) {
 		*ep++ = ATOM_UTF8_EXT;
-		put_int16(j, ep);
+		put_int16(len, ep);
 		ep += 2;
 	    }
 	    else {
 		*ep++ = SMALL_ATOM_UTF8_EXT;
-		put_int8(j, ep);
-		ep += 2;
+		put_int8(len, ep);
+		ep += 1;
 	    }
-	    sys_memcpy((char *) ep, (char*)atom_tab(i)->name, (int) j);
+	    sys_memcpy((char *) ep, (char *) a->name, len);
 	}
 	else {
-	    if (j <= 255 && (dflags & DFLAG_SMALL_ATOM_TAGS)) {
+	    if (a->latin1_chars <= 255 && (dflags & DFLAG_SMALL_ATOM_TAGS)) {
 		*ep++ = SMALL_ATOM_EXT;
-		j = erts_utf8_to_latin1(ep+1, atom_tab(i)->name, j);
-		put_int8(j, ep);
+		len = erts_utf8_to_latin1(ep+1, a->name, a->len);
+		ASSERT(len == a->latin1_chars);
+		put_int8(len, ep);
 		ep++;
 	    }
 	    else {
 		*ep++ = ATOM_EXT;
-		j = erts_utf8_to_latin1(ep+2, atom_tab(i)->name, j);
-		put_int16(j, ep);
+		len = erts_utf8_to_latin1(ep+2, a->name, a->len);
+		ASSERT(len == a->latin1_chars);
+		put_int16(len, ep);
 		ep += 2;
 	    }	    
 	}
-	ep += j;
+	ep += len;
 	return ep;
     }
 
@@ -1535,7 +1562,15 @@ dec_atom(ErtsDistExternal *edep, byte* ep, Eterm* objp)
                 goto error;
 	    }
         } else {
-            *objp = am_atom_put2(ep, len, is_latin1);
+	    Eterm atom = erts_atom_put(ep,
+				       len,
+				       (is_latin1
+					? ERTS_ATOM_ENC_LATIN1
+					: ERTS_ATOM_ENC_UTF8),
+				       0);
+	    if (is_non_value(atom))
+		goto error;
+            *objp = atom;
         }
 	ep += len;
 	break;
@@ -2248,7 +2283,15 @@ dec_term_atom_common:
 		    goto error;
 		}
 	    } else {
-	        *objp = am_atom_put2(ep, n, is_latin1);
+		Eterm atom = erts_atom_put(ep,
+					   n,
+					   (is_latin1
+					    ? ERTS_ATOM_ENC_LATIN1
+					    : ERTS_ATOM_ENC_UTF8),
+					   0);
+		if (is_non_value(atom))
+		    goto error;
+	        *objp = atom;
 	    }
 	    ep += n;
 	    break;
@@ -2917,18 +2960,22 @@ encode_size_struct2(ErtsAtomCacheMap *acmp, Eterm obj, unsigned dflags)
 		}
 	    }
 	    else {
-		int alen = atom_tab(atom_val(obj))->len;
-		result += 1 + 1 + alen;
-		if (dflags & DFLAG_UTF8_ATOMS) {
+		Atom *a = atom_tab(atom_val(obj));
+		int alen;
+		if ((dflags & DFLAG_UTF8_ATOMS) || a->latin1_chars < 0) {
+		    alen = a->len;
+		    result += 1 + 1 + alen;
 		    if (alen > 255) {
 			result++; /* ATOM_UTF8_EXT (not small) */
 		    }
-		    /*SVERK we use utf8 length which is an over estimation */
-		}		    
-		else if (alen > 255 || !(dflags & DFLAG_SMALL_ATOM_TAGS)) {
-		    result++; /* ATOM_EXT (not small) */
 		}
-		insert_acache_map(acmp, obj);
+		else {
+		    alen = a->latin1_chars;
+		    result += 1 + 1 + alen;
+		    if (alen > 255 || !(dflags & DFLAG_SMALL_ATOM_TAGS))
+			result++; /* ATOM_EXT (not small) */
+		}
+		insert_acache_map(acmp, obj, dflags);
 	    }
 	    break;
 	case SMALL_DEF: