diff --git a/src/ex_cmds.c b/src/ex_cmds.c
index b433bca..33021d2 100644
--- a/src/ex_cmds.c
+++ b/src/ex_cmds.c
@@ -5982,6 +5982,7 @@
     char_u	*line;
     int		in_example = FALSE;
     int		len;
+    char_u	*fname;
     char_u	*p;
     char_u	*rt;
     int		mustfree;
@@ -6028,124 +6029,187 @@
     }
 
     /*
-     * In the "help.txt" file, add the locally added help files.
-     * This uses the very first line in the help file.
+     * In the "help.txt" and "help.abx" file, add the locally added help
+     * files.  This uses the very first line in the help file.
      */
-    if (fnamecmp(gettail(curbuf->b_fname), "help.txt") == 0)
+    fname = gettail(curbuf->b_fname);
+    if (fnamecmp(fname, "help.txt") == 0
+#ifdef FEAT_MULTI_LANG
+	|| (fnamencmp(fname, "help.", 5) == 0
+	    && ASCII_ISALPHA(fname[5])
+	    && ASCII_ISALPHA(fname[6])
+	    && TOLOWER_ASC(fname[7]) == 'x'
+	    && fname[8] == NUL)
+#endif
+	)
     {
 	for (lnum = 1; lnum < curbuf->b_ml.ml_line_count; ++lnum)
 	{
 	    line = ml_get_buf(curbuf, lnum, FALSE);
-	    if (strstr((char *)line, "*local-additions*") != NULL)
+	    if (strstr((char *)line, "*local-additions*") == NULL)
+		continue;
+
+	    /* Go through all directories in 'runtimepath', skipping
+	     * $VIMRUNTIME. */
+	    p = p_rtp;
+	    while (*p != NUL)
 	    {
-		/* Go through all directories in 'runtimepath', skipping
-		 * $VIMRUNTIME. */
-		p = p_rtp;
-		while (*p != NUL)
+		copy_option_part(&p, NameBuff, MAXPATHL, ",");
+		mustfree = FALSE;
+		rt = vim_getenv((char_u *)"VIMRUNTIME", &mustfree);
+		if (fullpathcmp(rt, NameBuff, FALSE) != FPC_SAME)
 		{
-		    copy_option_part(&p, NameBuff, MAXPATHL, ",");
-		    mustfree = FALSE;
-		    rt = vim_getenv((char_u *)"VIMRUNTIME", &mustfree);
-		    if (fullpathcmp(rt, NameBuff, FALSE) != FPC_SAME)
-		    {
-			int	fcount;
-			char_u	**fnames;
-			FILE	*fd;
-			char_u	*s;
-			int	fi;
+		    int		fcount;
+		    char_u	**fnames;
+		    FILE	*fd;
+		    char_u	*s;
+		    int		fi;
 #ifdef FEAT_MBYTE
-			vimconv_T	vc;
-			char_u		*cp;
+		    vimconv_T	vc;
+		    char_u	*cp;
 #endif
 
-			/* Find all "doc/ *.txt" files in this directory. */
-			add_pathsep(NameBuff);
-			STRCAT(NameBuff, "doc/*.txt");
-			if (gen_expand_wildcards(1, &NameBuff, &fcount,
-					     &fnames, EW_FILE|EW_SILENT) == OK
-				&& fcount > 0)
-			{
-			    for (fi = 0; fi < fcount; ++fi)
-			    {
-				fd = mch_fopen((char *)fnames[fi], "r");
-				if (fd != NULL)
-				{
-				    vim_fgets(IObuff, IOSIZE, fd);
-				    if (IObuff[0] == '*'
-					    && (s = vim_strchr(IObuff + 1, '*'))
-								      != NULL)
-				    {
-#ifdef FEAT_MBYTE
-					int	this_utf = MAYBE;
-#endif
-					/* Change tag definition to a
-					 * reference and remove <CR>/<NL>. */
-					IObuff[0] = '|';
-					*s = '|';
-					while (*s != NUL)
-					{
-					    if (*s == '\r' || *s == '\n')
-						*s = NUL;
-#ifdef FEAT_MBYTE
-					    /* The text is utf-8 when a byte
-					     * above 127 is found and no
-					     * illegal byte sequence is found.
-					     */
-					    if (*s >= 0x80 && this_utf != FALSE)
-					    {
-						int	l;
-
-						this_utf = TRUE;
-						l = utf_ptr2len(s);
-						if (l == 1)
-						    this_utf = FALSE;
-						s += l - 1;
-					    }
-#endif
-					    ++s;
-					}
-#ifdef FEAT_MBYTE
-					/* The help file is latin1 or utf-8;
-					 * conversion to the current
-					 * 'encoding' may be required. */
-					vc.vc_type = CONV_NONE;
-					convert_setup(&vc, (char_u *)(
-						    this_utf == TRUE ? "utf-8"
-							  : "latin1"), p_enc);
-					if (vc.vc_type == CONV_NONE)
-					    /* No conversion needed. */
-					    cp = IObuff;
-					else
-					{
-					    /* Do the conversion.  If it fails
-					     * use the unconverted text. */
-					    cp = string_convert(&vc, IObuff,
-									NULL);
-					    if (cp == NULL)
-						cp = IObuff;
-					}
-					convert_setup(&vc, NULL, NULL);
-
-					ml_append(lnum, cp, (colnr_T)0, FALSE);
-					if (cp != IObuff)
-					    vim_free(cp);
+		    /* Find all "doc/ *.txt" files in this directory. */
+		    add_pathsep(NameBuff);
+#ifdef FEAT_MULTI_LANG
+		    STRCAT(NameBuff, "doc/*.??[tx]");
 #else
-					ml_append(lnum, IObuff, (colnr_T)0,
-								       FALSE);
+		    STRCAT(NameBuff, "doc/*.txt");
 #endif
-					++lnum;
-				    }
-				    fclose(fd);
+		    if (gen_expand_wildcards(1, &NameBuff, &fcount,
+					 &fnames, EW_FILE|EW_SILENT) == OK
+			    && fcount > 0)
+		    {
+#ifdef FEAT_MULTI_LANG
+			int	i1;
+			int	i2;
+			char_u	*f1;
+			char_u	*f2;
+			char_u	*t1;
+			char_u	*e1;
+			char_u	*e2;
+
+			/* If foo.abx is found use it instead of foo.txt in
+			 * the same directory. */
+			for (i1 = 0; i1 < fcount; ++i1)
+			{
+			    for (i2 = 0; i2 < fcount; ++i2)
+			    {
+				if (i1 == i2)
+				    continue;
+				if (fnames[i1] == NULL || fnames[i2] == NULL)
+				    continue;
+				f1 = fnames[i1];
+				f2 = fnames[i2];
+				t1 = gettail(f1);
+				if (fnamencmp(f1, f2, t1 - f1) != 0)
+				    continue;
+				e1 = vim_strrchr(t1, '.');
+				e2 = vim_strrchr(gettail(f2), '.');
+				if (e1 == NUL || e2 == NUL)
+				    continue;
+				if (fnamecmp(e1, ".txt") != 0
+				    && fnamecmp(e1, fname + 4) != 0)
+				{
+				    /* Not .txt and not .abx, remove it. */
+				    vim_free(fnames[i1]);
+				    fnames[i1] = NULL;
+				    continue;
+				}
+				if (fnamencmp(f1, f2, e1 - f1) != 0)
+				    continue;
+				if (fnamecmp(e1, ".txt") == 0
+				    && fnamecmp(e2, fname + 4) == 0)
+				{
+				    /* use .abx instead of .txt */
+				    vim_free(fnames[i1]);
+				    fnames[i1] = NULL;
 				}
 			    }
-			    FreeWild(fcount, fnames);
 			}
+#endif
+			for (fi = 0; fi < fcount; ++fi)
+			{
+			    if (fnames[fi] == NULL)
+				continue;
+			    fd = mch_fopen((char *)fnames[fi], "r");
+			    if (fd != NULL)
+			    {
+				vim_fgets(IObuff, IOSIZE, fd);
+				if (IObuff[0] == '*'
+					&& (s = vim_strchr(IObuff + 1, '*'))
+								  != NULL)
+				{
+#ifdef FEAT_MBYTE
+				    int	this_utf = MAYBE;
+#endif
+				    /* Change tag definition to a
+				     * reference and remove <CR>/<NL>. */
+				    IObuff[0] = '|';
+				    *s = '|';
+				    while (*s != NUL)
+				    {
+					if (*s == '\r' || *s == '\n')
+					    *s = NUL;
+#ifdef FEAT_MBYTE
+					/* The text is utf-8 when a byte
+					 * above 127 is found and no
+					 * illegal byte sequence is found.
+					 */
+					if (*s >= 0x80 && this_utf != FALSE)
+					{
+					    int	l;
+
+					    this_utf = TRUE;
+					    l = utf_ptr2len(s);
+					    if (l == 1)
+						this_utf = FALSE;
+					    s += l - 1;
+					}
+#endif
+					++s;
+				    }
+#ifdef FEAT_MBYTE
+				    /* The help file is latin1 or utf-8;
+				     * conversion to the current
+				     * 'encoding' may be required. */
+				    vc.vc_type = CONV_NONE;
+				    convert_setup(&vc, (char_u *)(
+						this_utf == TRUE ? "utf-8"
+						      : "latin1"), p_enc);
+				    if (vc.vc_type == CONV_NONE)
+					/* No conversion needed. */
+					cp = IObuff;
+				    else
+				    {
+					/* Do the conversion.  If it fails
+					 * use the unconverted text. */
+					cp = string_convert(&vc, IObuff,
+								    NULL);
+					if (cp == NULL)
+					    cp = IObuff;
+				    }
+				    convert_setup(&vc, NULL, NULL);
+
+				    ml_append(lnum, cp, (colnr_T)0, FALSE);
+				    if (cp != IObuff)
+					vim_free(cp);
+#else
+				    ml_append(lnum, IObuff, (colnr_T)0,
+								   FALSE);
+#endif
+				    ++lnum;
+				}
+				fclose(fd);
+			    }
+			}
+			FreeWild(fcount, fnames);
 		    }
-		    if (mustfree)
-			vim_free(rt);
 		}
-		break;
+		if (mustfree)
+		    vim_free(rt);
 	    }
+	    break;
 	}
     }
 }