From bf45e7ce36b16f2b78a76ce33801d806fb16ae22 Mon Sep 17 00:00:00 2001
From: "Reece H. Dunn" <msclrhd@gmail.com>
Date: Fri, 28 Apr 2017 07:08:02 +0100
Subject: [PATCH] tokenizer.c: Recognise U+0085 [NEW LINE (NEL)] as a newline
 codepoint.

---
 src/libespeak-ng/tokenizer.c |  9 +++++----
 tests/tokenizer.c            | 29 +++++++++++++++++++++++++++++
 2 files changed, 34 insertions(+), 4 deletions(-)

diff --git a/src/libespeak-ng/tokenizer.c b/src/libespeak-ng/tokenizer.c
index ced9f637..0abf0bb8 100644
--- a/src/libespeak-ng/tokenizer.c
+++ b/src/libespeak-ng/tokenizer.c
@@ -99,9 +99,10 @@ static espeakng_CTYPE codepoint_type(uint32_t c)
 
 	switch (c)
 	{
-	case '\r': return ESPEAKNG_CTYPE_CARRIAGE_RETURN;
-	case '\n': return ESPEAKNG_CTYPE_NEWLINE;
-	case '\0': return ESPEAKNG_CTYPE_END_OF_STRING;
+	case 0x0000: return ESPEAKNG_CTYPE_END_OF_STRING; // NULL
+	case 0x000A: return ESPEAKNG_CTYPE_NEWLINE; // LINE FEED (LF)
+	case 0x000D: return ESPEAKNG_CTYPE_CARRIAGE_RETURN; // CARRIAGE RETURN (CR)
+	case 0x0085: return ESPEAKNG_CTYPE_NEWLINE; // NEW LINE (NEL)
 	}
 
 	// 2. Classify codepoints by their Unicode General Category.
@@ -173,7 +174,7 @@ tokenizer_state_default(espeak_ng_TOKENIZER *tokenizer)
 			c = text_decoder_getc(tokenizer->decoder);
 		}
 		// fallthrough
-	case ESPEAKNG_CTYPE_NEWLINE: // '\n'
+	case ESPEAKNG_CTYPE_NEWLINE:
 		current += utf8_out(c, current);
 		*current = '\0';
 		return ESPEAKNG_TOKEN_NEWLINE;
diff --git a/tests/tokenizer.c b/tests/tokenizer.c
index 92216b13..2294fbe6 100644
--- a/tests/tokenizer.c
+++ b/tests/tokenizer.c
@@ -260,6 +260,34 @@ test_windows_newline_tokens()
 	destroy_tokenizer(tokenizer);
 }
 
+void
+test_unicode_newline_tokens()
+{
+	printf("testing unicode newline tokens\n");
+
+	espeak_ng_TOKENIZER *tokenizer = create_tokenizer();
+	espeak_ng_TEXT_DECODER *decoder = create_text_decoder();
+
+	assert(text_decoder_decode_string(decoder, "\xC2\x85\xC2\x85", -1, ESPEAKNG_ENCODING_UTF_8) == ENS_OK);
+	assert(tokenizer_reset(tokenizer, decoder) == 1);
+
+	// U+0085 : NEXT LINE (NEL) -- Used in EBCDIC systems as a combined CR+LF character.
+	assert(tokenizer_read_next_token(tokenizer) == ESPEAKNG_TOKEN_NEWLINE);
+	assert(tokenizer_get_token_text(tokenizer) != NULL);
+	assert(strcmp(tokenizer_get_token_text(tokenizer), "\xC2\x85") == 0);
+
+	assert(tokenizer_read_next_token(tokenizer) == ESPEAKNG_TOKEN_NEWLINE);
+	assert(tokenizer_get_token_text(tokenizer) != NULL);
+	assert(strcmp(tokenizer_get_token_text(tokenizer), "\xC2\x85") == 0);
+
+	assert(tokenizer_read_next_token(tokenizer) == ESPEAKNG_TOKEN_END_OF_BUFFER);
+	assert(tokenizer_get_token_text(tokenizer) != NULL);
+	assert(*tokenizer_get_token_text(tokenizer) == '\0');
+
+	destroy_text_decoder(decoder);
+	destroy_tokenizer(tokenizer);
+}
+
 void
 test_whitespace_tokens()
 {
@@ -340,6 +368,7 @@ main(int argc, char **argv)
 	test_linux_newline_tokens();
 	test_mac_newline_tokens();
 	test_windows_newline_tokens();
+	test_unicode_newline_tokens();
 	test_whitespace_tokens();
 
 	printf("done\n");