Patch to r102 of http://code.google.com/p/firefox-mac-pdf to support http://example.com/my.pdf#page=123 (jumps to page on load but doesn't respond to subsequent fragment edits)

minimal-page-fragment-support.patch

Index: src/PluginInstance.mm
===================================================================
--- src/PluginInstance.mm	(revision 102)
+++ src/PluginInstance.mm	(working copy)
@@ -185,8 +185,57 @@
   PDFDocument* document = [[[PDFDocument alloc] initWithData:pdfData] autorelease];
   [document setDelegate:self];
   [pdfView setDocument:document];
+  [self scrollToNewFragment:_url];
 }
 
+- (void)scrollToNewFragment:(NSString*)newURI
+{	
+  // Reference for fragment syntax is:
+  //   http://partners.adobe.com/public/developer/en/acrobat/PDFOpenParameters.pdf
+  // There's also the less featureful but slightly more standard RFC:
+  //   http://tools.ietf.org/html/rfc3778#section-3
+  // FIXME: support more general URI (without authority part) as well as URL.
+  NSString *fragment = [[NSURL URLWithString:newURI] fragment];
+  
+  // I *think* complete removal of a fragment has a larger-scale effect than
+  // just an onHashChange, so we don't need to skip to the top like hacking the
+  // fragment off an HTML page does.
+  if (fragment == nil) {
+    return;
+  }
+  NSArray *parameterList = [fragment componentsSeparatedByCharactersInSet:[NSCharacterSet characterSetWithCharactersInString:@"#&"]];
+  for (NSString *parameter in parameterList) {
+    NSArray *nameValuePair = [parameter componentsSeparatedByString:@"="];
+    if ([nameValuePair count] != 2) {
+      continue;
+    }
+    NSString *name = [nameValuePair objectAtIndex:0];
+    NSString *value = [nameValuePair objectAtIndex:1];
+
+    PDFDocument* document = [pdfView document];
+
+    // Only support page numbers at the moment (no nameddest etc).
+    if ([name isEqualToString:@"page"]) {
+      // PDFOpenParameters says "Specifies a numbered page in the document,
+      // using an integer value. The document\u2019s first page has a pagenum
+      // value of 1."  PDFView indices are zero-based.
+      //
+      // It'd be nice if the standard supported more abstraction --
+      // "#pagelabel=X" (or "#folio=X") -- to eliminate mental arithmetic when
+      // linking to front matter (e.g. "xi") or pages in the main text that are
+      // more than just integers (e.g. "2-1").
+      @try {
+        NSUInteger targetPageIndex = (NSUInteger)[value integerValue] - 1;
+        [pdfView goToPage:[document pageAtIndex:targetPageIndex]];
+      } @catch (NSException *exception) {
+        // Ignore malformed page number.
+      }
+    }
+    // Another useful abstraction might be "#annotation=box1".  Perhaps
+    // annotations could be included when '#search="term"' gets implemented?
+  }
+}
+
 - (void)_applyDefaults
 {
   bool autoScales = [Preferences getBoolPreference:"autoScales"];
Index: src/PluginInstance.h
===================================================================
--- src/PluginInstance.h	(revision 102)
+++ src/PluginInstance.h	(working copy)
@@ -57,6 +57,7 @@
 - (void)findPrevious;
 - (void)save;
 - (void)setData:(NSData*)data;
+- (void)scrollToNewFragment:(NSString*)newURI;
 - (void)setUrl:(NSString*)url;
 - (void)setVisible:(bool)visible;
 //- (void)loadURL:(NSString*)url;