From: Gerhard Sittig <redacted>
Date: Mon, 5 Jun 2017 16:35:22 +0000 (+0200)
Subject: input/csv: Skip leading UTF-8 BOM in the input stream
X-Git-Tag: libsigrok-0.5.0~15
X-Git-Url: https://sigrok.org/gitweb/?a=commitdiff_plain;h=4439363aa02928098eab8dcd18b5ccb41737c882;hp=ccff468b5e1b841a0c4f0426f502b9dad23dd9be;p=libsigrok.git

input/csv: Skip leading UTF-8 BOM in the input stream

This fixes bug #756.
---

diff --git a/src/input/csv.c b/src/input/csv.c
index 013d7a47..baf37188 100644
--- a/src/input/csv.c
+++ b/src/input/csv.c
@@ -613,6 +613,27 @@ out:
 	return ret;
 }
 
+/*
+ * Gets called from initial_receive(), which runs until the end-of-line
+ * encoding of the input stream could get determined. Assumes that this
+ * routine receives enough buffered initial input data to either see the
+ * BOM when there is one, or that no BOM will follow when a text line
+ * termination sequence was seen. Silently drops the UTF-8 BOM sequence
+ * from the input buffer if one was seen. Does not care to protect
+ * against multiple execution or dropping the BOM multiple times --
+ * there should be at most one in the input stream.
+ */
+static void initial_bom_check(const struct sr_input *in)
+{
+	static const char *utf8_bom = "\xef\xbb\xbf";
+
+	if (in->buf->len < strlen(utf8_bom))
+		return;
+	if (strncmp(in->buf->str, utf8_bom, strlen(utf8_bom)) != 0)
+		return;
+	g_string_erase(in->buf, 0, strlen(utf8_bom));
+}
+
 static int initial_receive(const struct sr_input *in)
 {
 	struct context *inc;
@@ -621,6 +642,8 @@ static int initial_receive(const struct sr_input *in)
 	char *p;
 	const char *termination;
 
+	initial_bom_check(in);
+
 	inc = in->priv;
 
 	termination = get_line_termination(in->buf);