When starting a new input line, even when continuing the same output

line, use the current output position as the reference position
for tabs on that input line.  This brings mandoc in line with the
behaviour of GNU, Heirloom, and Plan 9 roff.

1 files changed, 37 insertions, 6 deletions

diff --git a/term.c b/term.c
index 9823bb92..74d9db67 100644
--- a/term.c
+++ b/term.c
@@ -157,7 +157,11 @@ term_flushln(struct termp *p)
 		/* Finally, print the field content. */
 
 		term_field(p, vbl, nbr);
-		p->tcol->taboff += vbr + (*p->width)(p, ' ');
+		if (vbr < vtarget)
+			p->tcol->taboff += vbr;
+		else
+			p->tcol->taboff += vtarget;
+		p->tcol->taboff += (*p->width)(p, ' ');
 
 		/*
 		 * If there is no text left in the field, exit the loop.
@@ -177,7 +181,9 @@ term_flushln(struct termp *p)
 					vbr += (*p->width)(p, ' ');
 				continue;
 			case '\n':
+			case ASCII_NBRZW:
 			case ASCII_BREAK:
+			case ASCII_TABREF:
 				continue;
 			default:
 				break;
@@ -258,9 +264,11 @@ term_fill(struct termp *p, size_t *nbr, size_t *vbr, size_t vtarget)
 	size_t	 vn;        /* Visual position of the next character. */
 	int	 breakline; /* Break at the end of this word. */
 	int	 graph;     /* Last character was non-blank. */
+	int	 taboff;    /* Temporary offset for literal tabs. */
 
 	*nbr = *vbr = vis = 0;
 	breakline = graph = 0;
+	taboff = p->tcol->taboff;
 	for (ic = p->tcol->col; ic < p->tcol->lastcol; ic++) {
 		switch (p->tcol->buf[ic]) {
 		case '\b':  /* Escape \o (overstrike) or backspace markup. */
@@ -306,12 +314,19 @@ term_fill(struct termp *p, size_t *nbr, size_t *vbr, size_t vtarget)
 			*vbr = vis;
 			continue;
 
+		case ASCII_TABREF:
+			taboff = -vis - (*p->width)(p, ' ');
+			continue;
+
 		default:
 			switch (p->tcol->buf[ic]) {
 			case '\t':
-				vis += p->tcol->taboff;
+				if (taboff < 0 && (size_t)-taboff > vis)
+					vis = 0;
+				else
+					vis += taboff;
 				vis = term_tab_next(vis);
-				vis -= p->tcol->taboff;
+				vis -= taboff;
 				break;
 			case ASCII_NBRZW:  /* Non-breakable zero-width. */
 				break;
@@ -354,8 +369,10 @@ term_field(struct termp *p, size_t vbl, size_t nbr)
 	size_t	 vis;	/* Visual position of the current character. */
 	size_t	 vt;	/* Visual position including tab offset. */
 	size_t	 dv;	/* Visual width of the current character. */
+	int	 taboff; /* Temporary offset for literal tabs. */
 
 	vis = 0;
+	taboff = p->tcol->taboff;
 	for (ic = p->tcol->col; ic < nbr; ic++) {
 
 		/*
@@ -368,11 +385,17 @@ term_field(struct termp *p, size_t vbl, size_t nbr)
 		case ASCII_BREAK:
 		case ASCII_NBRZW:
 			continue;
+		case ASCII_TABREF:
+			taboff = -vis - (*p->width)(p, ' ');
+			continue;
 		case '\t':
 		case ' ':
 		case ASCII_NBRSP:
 			if (p->tcol->buf[ic] == '\t') {
-				vt = p->tcol->taboff + vis;
+				if (taboff < 0 && (size_t)-taboff > vis)
+					vt = 0;
+				else
+					vt = vis + taboff;
 				dv = term_tab_next(vt) - vt;
 			} else
 				dv = (*p->width)(p, ' ');
@@ -437,10 +460,10 @@ endline(struct termp *p)
 void
 term_newln(struct termp *p)
 {
-	p->tcol->taboff = 0;
 	p->flags |= TERMP_NOSPACE;
 	if (p->tcol->lastcol || p->viscol)
 		term_flushln(p);
+	p->tcol->taboff = 0;
 }
 
 /*
@@ -801,6 +824,14 @@ bufferc(struct termp *p, char c)
 		p->tcol->lastcol = p->col;
 }
 
+void
+term_tab_ref(struct termp *p)
+{
+	if (p->tcol->lastcol && p->tcol->lastcol <= p->col &&
+	    (p->flags & TERMP_NOBUF) == 0)
+		bufferc(p, ASCII_TABREF);
+}
+
 /*
  * See encode().
  * Do this for a single (probably unicode) value.
@@ -946,7 +977,7 @@ term_strlen(const struct termp *p, const char *cp)
 	const char	*seq, *rhs;
 	enum mandoc_esc	 esc;
 	static const char rej[] = { '\\', ASCII_NBRSP, ASCII_NBRZW,
-		ASCII_BREAK, ASCII_HYPH, '\0' };
+		ASCII_BREAK, ASCII_HYPH, ASCII_TABREF, '\0' };
 
 	/*
 	 * Account for escaped sequences within string length