GhostPDF - implement named Destinations

Bug #705218 "table of contents/bookmarks wrong after merging PDF files"

There are actually two problems here, this commit deals with the first
problem.

Destinations for Link annotations and Outlines need not be an array,
they can be a name or string and refer to a named destination. We need
to look up the name in the Dests name tree, which is stored in the
Names dictionary in the Catalog (Root) dictionary.

This deals with the fact that file1.pdf from this bug report has an
Outline with a Named Destination. We still need to deal with offsetting
the page numbers of the Destinations by the number of pages already
processed in previous files.

1 files changed, 100 insertions, 10 deletions

diff --git a/pdf/pdf_mark.c b/pdf/pdf_mark.c
index ac2de0a5d..51690dfcf 100644
--- a/pdf/pdf_mark.c
+++ b/pdf/pdf_mark.c
@@ -595,6 +595,82 @@ int pdfi_pdfmark_modDest(pdf_context *ctx, pdf_dict *link_dict)
     return code;
 }
 
+static int pdfi_get_named_dest(pdf_context *ctx, pdf_obj *Named, pdf_obj **Dest)
+{
+    int code = 0, len = 0;
+    pdf_dict *Names = NULL, *Dests = NULL;
+    pdf_array *NamesArray = NULL;
+    bool known;
+    char *str = NULL;
+
+    code = pdfi_dict_get_type(ctx, ctx->Root, "Names", PDF_DICT, (pdf_obj **)&Names);
+    if (code < 0)
+        goto error;
+
+    code = pdfi_dict_get_type(ctx, Names, "Dests", PDF_DICT, (pdf_obj **)&Dests);
+    if (code < 0)
+        goto error;
+
+    if (pdfi_type_of(Named) == PDF_NAME) {
+        code = pdfi_string_from_name(ctx, (pdf_name *)Named, &str, &len);
+        if (code < 0)
+            return code;
+    } else {
+        len = ((pdf_string *)Named)->length;
+        str = (char *)gs_alloc_bytes(ctx->memory, len, "pdfi_get_named_dest");
+        if (str == NULL) {
+            code = gs_note_error(gs_error_VMerror);
+            goto error;
+        }
+        memcpy(str, ((pdf_string *)Named)->data, len);
+        str[len] = 0;
+    }
+
+    /* At initial node, if we have a Names array here then this is the only array
+     * otherwise we will need to deal with Kids.
+     */
+    code = pdfi_dict_known(ctx, Dests, "Names", &known);
+    if (code < 0)
+        goto error;
+
+    if (known) {
+        int i = 0;
+        pdf_string *StrKey = NULL;
+
+        code = pdfi_dict_get_type(ctx, Dests, "Names", PDF_ARRAY, (pdf_obj **)&NamesArray);
+        if (code < 0)
+            goto error;
+        for (i = 0;i < pdfi_array_size(NamesArray); i+=2) {
+            code = pdfi_array_get_type(ctx, NamesArray, i, PDF_STRING, (pdf_obj **)&StrKey);
+            if (code < 0)
+                goto error;
+
+            if (StrKey->length == len && strncmp((const char *)StrKey->data, str, len) == 0) {
+                pdfi_countdown(StrKey);
+                code = pdfi_array_get(ctx, NamesArray, i + 1, (pdf_obj **)Dest);
+                break;
+            }
+            pdfi_countdown(StrKey);
+            StrKey = NULL;
+        }
+    } else {
+        code = pdfi_dict_get_type(ctx, Dests, "Kids", PDF_ARRAY, (pdf_obj **)&NamesArray);
+        if (code < 0)
+            goto error;
+
+    }
+
+error:
+    if (pdfi_type_of(Named) == PDF_NAME)
+        (void)pdfi_free_string_from_name(ctx, str);
+    else
+        gs_free_object(ctx->memory, str, "pdfi_get_named_dest");
+    pdfi_countdown(Names);
+    pdfi_countdown(Dests);
+    pdfi_countdown(NamesArray);
+    return code;
+}
+
 /* Special handling for "A" in Link annotations and Outlines
  * Will delete A if handled and if A_key is provided.
  */
@@ -635,19 +711,33 @@ int pdfi_pdfmark_modA(pdf_context *ctx, pdf_dict *dict)
     if (code <= 0) goto exit;
     /* We only handle GoTo for now */
     if (pdfi_name_is(S_name, "GoTo")) {
-        code = pdfi_dict_knownget_type(ctx, A_dict, "D", PDF_ARRAY, (pdf_obj **)&D_array);
-        if (code == 0) goto exit;
-        if (code < 0) {
-            if (code == gs_error_typecheck) {
-                /* TODO: Are there other cases to handle?
-                 * Sample tests_private/pdf/sumatra/recursive_action_destinations.pdf
-                 * has a recursive destination that has an indirect ref here.  We return a
-                 * typecheck and that causes us to omit the whole thing, but is that
-                 * really the best treatment?
-                 */
+        code = pdfi_dict_knownget(ctx, A_dict, "D", (pdf_obj **)&D_array);
+        if (code <= 0)
+            goto exit;
+        if (pdfi_type_of(D_array) == PDF_STRING || pdfi_type_of(D_array) == PDF_NAME)
+        {
+            pdf_obj *Dest = NULL;
+
+            code = pdfi_get_named_dest(ctx, (pdf_obj *)D_array, &Dest);
+            if (code < 0)
+                goto exit;
+            pdfi_countdown(D_array);
+            D_array = NULL;
+            if (pdfi_type_of(Dest) != PDF_DICT) {
+                pdfi_countdown(Dest);
+                code = gs_note_error(gs_error_typecheck);
+                goto exit;
             }
+            code = pdfi_dict_knownget(ctx, (pdf_dict *)Dest, "D", (pdf_obj **)&D_array);
+            pdfi_countdown(Dest);
+            if (code <= 0)
+                goto exit;
+        }
+        if (pdfi_type_of(D_array) != PDF_ARRAY) {
+            code = gs_note_error(gs_error_typecheck);
             goto exit;
         }
+
         /* Process the D array to replace with /Page /View */
         code = pdfi_pdfmark_add_Page_View(ctx, dict, D_array);
         if (code < 0) goto exit;