The Halibut manual contained at least one instance of two index

terms (intentionally) differing only in case, which were being
silently folded into one by the case-insensitive index tag
comparison. Halibut now warns in this situation (but then folds them
anyway, which I think is better than silently generating an index
containing many case-distinct forms of the same word - I imagine
it's very easy to do that by mistake). The manual has been fixed to
explicitly define distinct keywords (in the case I spotted and in
five other cases picked up by the new warning!), and also documents
this issue and how to work with it.

[originally from svn r4279]

1 files changed, 13 insertions, 2 deletions

diff --git a/index.c b/index.c
index 5dde59f..628934e 100644
--- a/index.c
+++ b/index.c
@@ -109,8 +109,6 @@ void index_merge(indexdata *idx, int is_explicit, wchar_t *tags, word *text,
 	    t->implicit_text = text;
 	    t->implicit_fpos = *fpos;
 	} else {
-	    sfree(t);
-	    t = existing;
 	    if (!is_explicit) {
  		/*
 		 * An implicit \IM for a tag that's had an implicit
@@ -119,12 +117,25 @@ void index_merge(indexdata *idx, int is_explicit, wchar_t *tags, word *text,
 		 * differences. And check the tag for case match
 		 * against the existing tag, likewise.
 		 */
+
+		/*
+		 * Check the tag against its previous occurrence to
+		 * see if the cases match.
+		 */
+		if (ustrcmp(t->name, existing->name)) {
+		    error(err_indexcase, fpos, t->name,
+			  &existing->implicit_fpos, existing->name);
+		}
+
+		sfree(t);
 	    } else {
 		/*
 		 * An explicit \IM added to a valid tag. In
 		 * particular, this removes the implicit \IM if
 		 * present.
 		 */
+		sfree(t);
+		t = existing;
 		if (t->implicit_text) {
 		    free_word_list(t->implicit_text);
 		    t->implicit_text = NULL;