TodoList

0.5.2 Bugfixes

• [052] While a menu is selected, I type a character in order to set the menu accelerator. There is a crash at this point. Although, this could be a glade bug (and I think that it is).... (Nathan)
  • We might have to move all of the menu handling code into the C source files to fix this. I don't think that Glade can do it correctly.
• [052, MINOR] I'm not sure how important this is...

 UPDATE: waiting for request...
 UPDATE: processing received request (http://www.kempa.com/blog/index.rdf)
 UPDATE: downloading http://www.kempa.com/blog/index.rdf

 Program received signal SIGSEGV, Segmentation fault.
 [Switching to Thread 1083725312 (LWP 4731)]
 0x4052c4b4 in xmlStrEqual () from /usr/lib/libxml2.so.2
 (gdb) bt
 #0  0x4052c4b4 in xmlStrEqual () from /usr/lib/libxml2.so.2
 #1  0x404f7f96 in htmlEntityLookup () from /usr/lib/libxml2.so.2
 #2  0x404f9202 in htmlParseEntityRef () from /usr/lib/libxml2.so.2
 #3  0x404fb2ce in htmlParseCharRef () from /usr/lib/libxml2.so.2
 #4  0x404fcda2 in htmlCreateMemoryParserCtxt () from /usr/lib/libxml2.so.2
 #5  0x404fd37f in htmlParseChunk () from /usr/lib/libxml2.so.2
 #6  0x08057e42 in unhtmlize (string=0xaebff50 "  #7  0x080807a1 in parseRSSItem (fp=0x6e, cp=0x6e, cur=0x0) at rss_item.c:259

• [052] popup windows for deleted feeds cause a crash when they are clicked...
• some favicons are totally black. example are salon.com (32x32 24bit) and Sam Ruby's blog
  • the salon cache file is a single vertical black bar 16 pixel high and with various width -> gdkpixbuf scaling problem? (Lars)
• [052] some of the last changes to gtkhtml2 must have broke the text popup menu handling, one always get the link popup menu
• [052, IN WORK] implement new cache model

• [052] Implement the unhtmlize(), unescape(), g_markup_escape_text() solution to fix markup in item and feed titles. (Lars will code)
  • Also, I think that bits of cleanup should be done:
    • Remove all trailing and leading whitespace.
    • Replace all instances of multiple whitespace in a row with a single space. (for example, remove newlines).
    • And finaly, replace <BR> tags with newlines.
  • Should be done in two places. There should be a generic HTML stripping routine and a special title treatment like Lucas Nussbaum proposed. (Lars)
  • Maybe the feed titles should be run through the filter only during subscribing because they are changable by the user. The user might want to put spaces in the front of the title, etc.... But, the item titles should be run through the filter ony when displayed.
• [052] Change action of "Update feeds on startup" to if it is disabled, reset all poll times of feeds after loading. Does this make sense? The 'smart-update' on startup works very well for me. (Nathan)
  • I also think this option is now obselete. Good idea. (Lars)
  • But, what is a good name for this option?
  • Maybe something like "don't update outdated feeds on startup" (Lars)
• [052] Implement solution to avoid multiple running instances
• XML namespaces should be checked using the URL of the namespace, not the actual namespace tag used. See "http://www.xmlsoft.org/namespaces.html"
  • Why? What is the advantage? (Lars)
  • If a feed uses a namespace as the default, then it won't have a prefix. Also, the prefixs can be set to anything the feed author wants. (Nathan)
  • True, but you will never see this in the field as long as RSS is not XML. So when the standard defines rdf:items and you write items (along with default NS) 90% of all feed readers 'll fail. Would be good to have if it is simple to implement. (Lars)
  • It should be simple. Have a second hashmap of url->ns handlers. First try to search for the handler via the urlHashMap, if that fails fall back on the namespace name handler. cur_ptr->nsUrl is where the URL is located IIRC. (Nathan)
    • Ok. But I don't want to invest my time in this :-) (Lars)
  • This might go well with a redesign of how feed/item metadata is stored internally that we discussed. (Nathan)
    • Yes. We should at least decide on the metadata solution to avoid any incompatibilities. (Lars)
• [052] fix the problem that one can't enter spaces into HTML input elements, the space hotkey should only work when the html view is not focussed
• [052-053] Finish VFolder design

• Support skipping of already subscribed feeds when importing from OPML.
• Show email adresses in the meta information of Atom or RSS as email links.
• [060-06x] Implement VFolder support.
• Add command line options to suppress functions like update on startup or load a specific HTML widget to allow users to workaround crashes.
• maybe write an RDF parser to read RDF metadata (like a set of dc:contributors)
• check our OPML syntax against this DTD from Danny Ayers http://semtext.org/wiki/index.php/OpmlDtd (he does not has htmlUrl but url and uses text instead title, which does not match most aggregators behaviour)
  • We use a bunch of extra tags, so we would need to make our own DTD/Schema.(Nathan)
  • Not sure. I understand OPML (and XML in general) this way that you follow to XML application spec (in this case OPML), which means you support all mandatory tags and use optional tags in the right way. And you add new own tags as you want, but you don't expect that anyone reads them (so no DTD is needed). (Lars)
• Adding an OPML file with folders to the feedlist causes the folders to be displayed as headlines, and not the feeds that they contain. How about making directories put their contents directly in the feedlist?
• Find a way to set colors of the cell renderers without having problems with GTK themes that uses dark or black background, which makes the feedlist invisible
• Liferea for GNOME see http://robert.techcommons.info/archives/2004/05/06/reui/
• some more test feeds: http://xoomer.virgilio.it/dialtone/out.py
• [maybe 0.5.2-0.5.3] rewrite caching to fix localization problems and memory consumption
• Mozilla does not reset scrolling when selecting "next unread item". It does when manuallly selecting... (Lars)
• All GConf? code should use notification callbacks. Also, provide a gconf schema file.
• Advanced favicon detection based on parsing the HTML version of the feed, and then defaulting to guessing.
• [090] remove all migration code from earlier releases (cache migration, feedlist migration...)

• [052, FIXED] "This feed specifies no default update interval" is untranslatable.
• [052, FIXED] Passing URL's with '#' characters does not work. Currently these are incorrectly escaped as %23.
• [052, DONE] maybe update favicons after some time (one month)
• [052, DONE] We should propagate the proxy settings to Mozilla, maybe reuse the Moz. preferences code from Galeon
• [052, FIXED] Sam Ruby pointed out a number of aggregators including Liferea that do not respect HTTP 410
  • see http://intertwingly.net/stories/2004/06/13/gone
• [052, DONE] redesign and redocument cache model localization problems and memory consumption
  • see MetadataDesign

OldTodoList