X-Git-Url: https://git.distorted.org.uk/~mdw/sgt/halibut/blobdiff_plain/41848b3e209f21729cbfe29aaa47d4c061aed94e..39d3d9002fc5eb3a66e926af9e5350030762add9:/input.c

diff --git a/input.c b/input.c
index a616ac9..337517d 100644
--- a/input.c
+++ b/input.c
@@ -426,6 +426,7 @@ token get_token(input *in) {
 	    c == '#' || c == '{' || c == '}' || c == '.') {
 	    /* single-char command */
 	    rdadd(&rs, c);
+	    prevpos = rsc.pos;
 	} else if (c == 'u') {
 	    int len = 0;
 	    do {
@@ -609,7 +610,7 @@ static void read_file(paragraph ***ret, input *in, indexdata *idx,
 	    stack_style = 2,	       /* \e, \c, \cw */
 	    stack_idx = 4,	       /* \I, \i, \ii */
 	    stack_hyper = 8,	       /* \W */
-	    stack_quote = 16,	       /* \q */
+	    stack_quote = 16	       /* \q */
 	} type;
 	word **whptr;		       /* to restore from \u alternatives */
 	word **idximplicit;	       /* to restore from \u alternatives */
@@ -902,7 +903,21 @@ static void read_file(paragraph ***ret, input *in, indexdata *idx,
 		/* Get keywords. */
 		dtor(t), t = get_token(in);
 		fp = t.pos;
-		while (t.type == tok_lbrace) {
+		while (t.type == tok_lbrace ||
+		       (t.type == tok_white && (needkw & 24))) {
+		    /*
+		     * In paragraph types which can't accept any
+		     * body text (such as \cfg), we are lenient
+		     * about whitespace between keywords. This is
+		     * important for \cfg in particular since it
+		     * can often have many keywords which are long
+		     * pieces of text, so it's useful to permit the
+		     * user to wrap the line between them.
+		     */
+		    if (t.type == tok_white) {
+			dtor(t), t = get_token(in); /* eat the space */
+			continue;
+		    }
 		    /* This is a keyword. */
 		    nkeys++;
 		    /* FIXME: there will be bugs if anyone specifies an
@@ -1528,7 +1543,17 @@ static void read_file(paragraph ***ret, input *in, indexdata *idx,
 	}
 	stk_free(parsestk);
 	prev_para_type = par.type;
-	addpara(par, ret);
+	/*
+	 * Before we add the paragraph to the output list, we
+	 * should check that there was any text in it at all; there
+	 * might not be if (for example) the paragraph contained
+	 * nothing but an unrecognised command sequence, and if we
+	 * put an empty paragraph on the list it may confuse the
+	 * back ends later on.
+	 */
+	if (par.words) {
+	    addpara(par, ret);
+	}
 	if (t.type == tok_eof)
 	    already = TRUE;
     }
@@ -1550,10 +1575,26 @@ static void read_file(paragraph ***ret, input *in, indexdata *idx,
     stk_free(crossparastk);
 }
 
+struct {
+    char const *magic;
+    size_t nmagic;
+    void (*reader)(input *);
+} magics[] = {
+    { "%!FontType1-",     12, &read_pfa_file },
+    { "%!PS-AdobeFont-",  15, &read_pfa_file },
+    { "\x80\x01",          2, &read_pfb_file },
+    { "StartFontMetrics", 16, &read_afm_file },
+    { "\x00\x01\x00\x00",  4, &read_sfnt_file },
+    { "true",		   4, &read_sfnt_file },
+};
+
 paragraph *read_input(input *in, indexdata *idx) {
     paragraph *head = NULL;
     paragraph **hptr = &head;
     tree234 *macros;
+    char mag[16];
+    size_t len, i;
+    void (*reader)(input *);
 
     macros = newtree234(macrocmp);
 
@@ -1565,7 +1606,20 @@ paragraph *read_input(input *in, indexdata *idx) {
 	    in->csstate = charset_init_state;
 	    in->wcpos = in->nwc = 0;
 	    in->pushback_chars = NULL;
-	    read_file(&hptr, in, idx, macros);
+	    reader = NULL;
+	    len = fread(mag, 1, sizeof(mag), in->currfp);
+	    for (i = 0; i < lenof(magics); i++) {
+		if (len >= magics[i].nmagic &&
+		    memcmp(mag, magics[i].magic, magics[i].nmagic) == 0) {
+		    reader = magics[i].reader;
+		    break;
+		}
+	    }
+	    rewind(in->currfp);
+	    if (reader == NULL)
+		read_file(&hptr, in, idx, macros);
+	    else
+		(*reader)(in);
 	}
 	in->currindex++;
     }