From 328e34077a2017d4017f907ac69c7ed5d6c160d4 Mon Sep 17 00:00:00 2001
From: David Phillips <david@sighup.nz>
Date: Wed, 1 Aug 2018 20:21:37 +1200
Subject: Trim leading whitespace from contextual error, detect EOF

Side-note that the token location tuple maintains the correct column
number to include the whitespace in the source file. Side-side note:
tabs are counted as one column but many editors will count them as
whatever the tabstop/tabwidth is.
---
 common.c |  8 +++++++-
 lexer.c  |  7 +++++++
 lexer.h  |  1 +
 parser.c | 30 ++++++++++++++++++++++++------
 4 files changed, 39 insertions(+), 7 deletions(-)

diff --git a/common.c b/common.c
index 96c9b72..c0e0af4 100644
--- a/common.c
+++ b/common.c
@@ -7,13 +7,19 @@ indicate_file_area(FILE* fd, size_t line, size_t column, size_t span) {
 	char margin[] = "  ";
 	/* FIXME use proper line counting, not this hack */
 	char buf[1024];
+	char *line_start = &buf;
+
 	rewind(fd);
 	for (; line; line--) {
 		fgets(buf, sizeof(buf), fd);
 	}
 
+	while (*line_start == '\t' || *line_start == ' ') {
+		line_start++;
+	}
+
 	fputs(margin, stderr);
-	fputs(buf, stderr);
+	fputs(line_start, stderr);
 
 	/* corner case (still needed?) - buf was just return */
 	if (strlen(buf) == 1 && buf[0] == '\n') {
diff --git a/lexer.c b/lexer.c
index a8b7716..d9810a5 100644
--- a/lexer.c
+++ b/lexer.c
@@ -53,6 +53,7 @@ static char buf[BUFFER_SIZE];
 static FILE* fd;
 static size_t line_number = 0;
 static size_t column_number = 0;
+static size_t leading_whitespace_len = 0;
 static struct token *tok_start = NULL;
 static struct token *tok_cursor = NULL;
 
@@ -64,6 +65,7 @@ get_current_loc(void) {
 	struct location l;
 	l.line = line_number;
 	l.column = column_number + 1;
+	l.leading_whitespace_len = leading_whitespace_len;
 	return l;
 }
 
@@ -187,6 +189,8 @@ lex_colon(void) {
 static int
 lex_line(void) {
 	size_t length = strlen(buf);
+	leading_whitespace_len = 0;
+
 	while (column_number < length && column_number < strlen(buf)) {
 		switch (buf[column_number]) {
 			case ':':
@@ -195,6 +199,9 @@ lex_line(void) {
 			case ' ':
 			case '\t':
 				eat_whitespace();
+				if (leading_whitespace_len == 0) {
+					leading_whitespace_len = column_number;
+				}
 				break;
 			case '\r':
 			case '\n':
diff --git a/lexer.h b/lexer.h
index 2e27ffa..3ce15f4 100644
--- a/lexer.h
+++ b/lexer.h
@@ -19,6 +19,7 @@ enum TOKEN_TYPE {
 struct location {
 	size_t line;
 	size_t column;
+	size_t leading_whitespace_len;
 };
 
 struct token {
diff --git a/parser.c b/parser.c
index 618df1a..f16af6c 100644
--- a/parser.c
+++ b/parser.c
@@ -14,9 +14,16 @@ static FILE* fd;
 static struct token *cursor;
 static const char *filename;
 
-#define emit(...) fprintf(stderr, "%s at (%zd,%zd): ", filename, cursor->loc.line, cursor->loc.column);\
-                  fprintf(stderr, __VA_ARGS__);\
-                  indicate_file_area(fd, cursor->loc.line, cursor->loc.column, cursor->span)
+/* FIXME change to varadic fuction */
+#define emit(...) \
+    if (cursor) { \
+        fprintf(stderr, "%s at (%zd,%zd): ", filename, cursor->loc.line, cursor->loc.column); \
+        fprintf(stderr, __VA_ARGS__); \
+        indicate_file_area(fd, cursor->loc.line, cursor->loc.column - cursor->loc.leading_whitespace_len, cursor->span); \
+    } else { \
+        fprintf(stderr, "%s: ", filename); \
+        fprintf(stderr, __VA_ARGS__); \
+    }
 
 
 //static struct op_lookup uop_handlers[] = {
@@ -42,9 +49,13 @@ expect(enum TOKEN_TYPE e) {
 	char *expected_desc = "(internal error)";
 	char *observed_desc = "(internal error)";
 
-	if (cursor->type != e) {
+	if (!cursor || cursor->type != e) {
 		expected_desc = get_token_description(e);
-		observed_desc = get_token_description(cursor->type);
+		if (cursor) {
+			observed_desc = get_token_description(cursor->type);
+		} else {
+			observed_desc = "end of file";
+		}
 		emit("Error: Expected %s, got %s\n", expected_desc, observed_desc);
 		return 1;
 	}
@@ -54,7 +65,9 @@ expect(enum TOKEN_TYPE e) {
 
 static void
 kerchunk() {
-	cursor = cursor->next;
+	if (cursor) {
+		cursor = cursor->next;
+	}
 }
 
 int
@@ -64,6 +77,11 @@ parse(const char *fname, FILE *f, struct token *t) {
 	filename = fname;
 	cursor = t;
 
+	/* Eat leading EOL tokens */
+	while (cursor && cursor->type == TOK_EOL) {
+		kerchunk();
+	}
+
 	if (expect(TOK_MODULE)) {
 		return 1;
 	}
-- 
cgit v1.1