Ben Hutchings points out that `.UC', which I think I must have

[sgt/halibut] / input.c

diff --git a/input.c b/input.c
index 365fec0..d8d44f9 100644 (file)
--- a/input.c
+++ b/input.c
@@ -103,6 +103,8 @@ static int get(input *in, filepos *pos, rdstringc *rsc) {
     }
     else if (in->stack) {
        wchar_t c = in->stack->text[in->stack->ptr];
+        if (pos)
+            *pos = in->stack->pos;
        if (in->stack->text[++in->stack->ptr] == L'\0') {
            macrostack *tmp = in->stack;
            in->stack = tmp->next;
@@ -424,6 +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 {
@@ -589,11 +592,11 @@ static paragraph *addpara(paragraph newpara, paragraph ***hptrptr) {
 /*
  * Reads a single file (ie until get() returns EOF)
  */
-static void read_file(paragraph ***ret, input *in, indexdata *idx) {
+static void read_file(paragraph ***ret, input *in, indexdata *idx,
+                     tree234 *macros) {
     token t;
     paragraph par;
     word wd, **whptr, **idximplicit;
-    tree234 *macros;
     wchar_t utext[2], *wdtext;
     int style, spcstyle;
     int already;
@@ -629,7 +632,6 @@ static void read_file(paragraph ***ret, input *in, indexdata *idx) {
 
     t.text = NULL;
     t.origtext = NULL;
-    macros = newtree234(macrocmp);
     already = FALSE;
 
     crossparastk = stk_new();
@@ -806,6 +808,11 @@ static void read_file(paragraph ***ret, input *in, indexdata *idx) {
            continue;
        }
 
+       while (t.type == tok_cmd &&
+              macrolookup(macros, in, t.text, &t.pos)) {
+           dtor(t), t = get_token(in);
+       }
+
        /*
         * This token begins a paragraph. See if it's one of the
         * special commands that define a paragraph type.
@@ -1522,7 +1529,19 @@ 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);
+       } else {
+           error(err_emptypara, &par.fpos);
+       }
        if (t.type == tok_eof)
            already = TRUE;
     }
@@ -1540,7 +1559,6 @@ static void read_file(paragraph ***ret, input *in, indexdata *idx) {
      * this cleanup doesn't happen.
      */
     dtor(t);
-    macrocleanup(macros);
 
     stk_free(crossparastk);
 }
@@ -1548,6 +1566,9 @@ static void read_file(paragraph ***ret, input *in, indexdata *idx) {
 paragraph *read_input(input *in, indexdata *idx) {
     paragraph *head = NULL;
     paragraph **hptr = &head;
+    tree234 *macros;
+
+    macros = newtree234(macrocmp);
 
     while (in->currindex < in->nfiles) {
        in->currfp = fopen(in->filenames[in->currindex], "r");
@@ -1557,10 +1578,12 @@ 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);
+           read_file(&hptr, in, idx, macros);
        }
        in->currindex++;
     }
 
+    macrocleanup(macros);
+
     return head;
 }