Permalink
Switch branches/tags
Nothing to show
Commits on Mar 2, 2012
  1. Key cleanup.

    committed Mar 2, 2012
Commits on Aug 4, 2011
  1. Stop getting 'disconnected' messages when we aren't actually disconne…

    …cted.
    
    After poking around, it looks like the lag between heartbeats is hovering at around 8-12 seconds. I found this by poking around in the Chrome console:
    
      (new Date().getTime()/1000) - PissWhistle.connection.latestHeartbeat
    
    Running that repeatedly showed that the time was greater than the acceptable lag.
    
    I think this could be because of differences in the time between the server (which sends the heartbeat) and the client (which compares that time to its local time), but I'm not sure of that.
    
    Regardless, even with the disconnected banner visible I could still reliably post messages.
    committed Aug 4, 2011
Commits on Jun 2, 2011
  1. Don't redraw all presence messages. Only add/remove individual users …

    …so that status message isn't lost.
    
    This change now means that multiple presences for a single user won't be displayed because of the way it's checking who's already connected to the stream.
    kalv committed Jun 2, 2011
Commits on May 31, 2011
  1. Explicitly request json data.

    committed May 31, 2011
Commits on May 26, 2011
Commits on May 9, 2011
  1. Latching into the disconnected state and never checking the connectio…

    …n again was too strict.
    
    By setting up a new timeout in each call to check_connection (whatever the state of the connection), we avoid the latching behaviour.
    
    The return value from PissWhistle.check_connection() seemed to be the opposite of what I would expect and since it doesn't appear to be used anywhere, I've removed it.
    floehopper committed May 9, 2011
Commits on Apr 11, 2011
Commits on Mar 28, 2011
Commits on Mar 23, 2011
  1. Historical chat messages weren't being handled correctly in Safari.

    This is because they were using Safari's stricter Date parsing directly; by extracting a function from our own iso8601 function, we can fix this.
    
    We decided to implement this with a chainy-object thing to clarify the purpose of each of the steps. Only time can judge us.
    committed Mar 23, 2011
  2. Re-organise Javascript so that sprockets stops being appended to and …

    …also to stop sinatra serving the file.
    
    The Javascript source files have been moved to the lib folder, only sprockets will be served from public/javascripts.
    kalv committed Mar 23, 2011
  3. The iso8601 only accepts datetimes as strings, but thankfully now it …

    …can deal with the pushed_at strings from github itself.
    committed Mar 23, 2011
  4. A more robust fix for Safari's date handling.

    Yeah, I should've written tests. This is clear to me now. We've tested the most common formats of date that we could find in the message data, along with one format that we use in another test; date processing seems to work in both Chrome and Safari now.
    committed Mar 23, 2011
  5. Commit 505e0fc vendorised the third party javascript libraries which …

    …stopped the tests from running.
    committed Mar 23, 2011
  6. Handle an error I was seeing in Firefox 3 where null data is returned…

    … from the HTTP POST to get an access token.
    floehopper committed Mar 23, 2011
  7. Safari (and therefore Fluid) cannot create Date objects with dates th…

    …at contain hyphens, Ts or Zs
    committed Mar 23, 2011
Commits on Mar 21, 2011
  1. Word wrap pastes please so that I don't have to copy and paste it out…

    … again to read it
    kalv committed Mar 21, 2011
Commits on Mar 17, 2011
Commits on Mar 16, 2011
  1. Apparently it's not as hard to parse this as I thought.

    Hurrah for no broken tests.
    committed Mar 16, 2011
  2. It's not necessary for Debug to inherit properties from Base.

    Clarifying which handlers use functionality from Base, and which don't, will make refactoring easier in the future.
    committed Mar 16, 2011
  3. This code isn't actively used.

    If we ever need it, it's always here in Git-land.
    committed Mar 16, 2011
  4. This function isn't used.

    It's probably useful to have something like it in the future, but for the moment lets keep the code clean.
    committed Mar 16, 2011
  5. ... and use idiomatic naming.

    committed Mar 16, 2011
  6. Use an 'infinite history' scroll.

    This further encapsulates chat-specific behaviour in the handler, and out of the main library.
    committed Mar 16, 2011
  7. By making the loadMessages function more generic, we don't need the l…

    …oadHistory function at all.
    committed Mar 16, 2011
  8. The Chat handler process function has enough data in the messages (in…

    … the DOM) and the new message data to determine whether to invoke the 'historical' behaviour.
    
    This avoids the caller of the Chat handler process function needing to know whether or not the 'historical' behaviour should be invoked.
    lazyatom-and-floehopper committed Mar 16, 2011