Preservation | Intimidation


Here it is, quick and dirty. What have we, the fast and the furious, changed in this here website.

[% return "May 1, 2002" %]

  • formxml_usergroup: XMLTrue now supports usergroups. It does support links_noparse, but usergroups exec like superdocs, so the output is probably not reliable (see the edev node for an example). It also includes the weblog information. You may want to exclude the description by tacking on no_descrip=1 to the URL. You'll also get a weblog ONLY if you are a member of that usergroup. This was requested by fuzzie.
  • Updated: Honor Roll and you to um, actually finish it. Created level devotion in the process. Also altered A glimpse ahead so that pp can check the numbers before this next phase comes out of beta.

[% return "May 2, 2002" %]

  • clientdev API change: all exported writeups now add an embedded e2link for the <parent> node. The change goes through xmlwriteup;
  • Fixed up A Glimpse Ahead and Honor Roll and You to remove several spelling / date errors. Also fixed a logic error (affecting tbbk), that made it possible to have xp and devotion levels of 0. Whoops.
  • Fixed a possible server error in Scratch Pad Viewer where we weren't loading Everything::Experience; Fixed and working now. Thanks to abiesseu for reporting it.
  • Made consistency fixes to statssection_personal, a glimpse ahead, and Honor roll and you as requested by professor pi. Made sure devotion is an even number and merit is "%.2lf"
  • Core update:

    Okay kids, get comfy. Let me tell you a story.

    "A stooooory uncle jay?"

    Yes, a story

    *rustling as children settle in, eyes focusing on the weary developer*

    So, we've had some core bugs we've fixed today. A long time ago, in a land farrr farrr away... Hey, can you name something that happened a long time ago? How about you little Suzie:

    "Starr Wars Epimisode One"

    Yes little Suzie, but older than that

    eyes wide crowd: "WOW!"

    Yes, even as old as E2! Did you know...

    • That if you used a URL string of a node name and a lastnode_id that it wouldn't leave a softlink if there was more than one item and you didn't have read access to the second item? Bizarre. Normally it would go through the Duplicates found page, and you'd pick, but I guess not. This is because we'd call displayPage instead of gotoNode. The same basic bug also manifested itself as:
    • (and this was recent) If you tried to softlink to a node that also had a user, you wouldn't be able to because we called displayPage instead of gotoNode.
    • If you tried to edit an e2node with an edit displaytype, you'd always see the edit page, whether or not you could normally edit it, and that was because we were short circuiting the code inside of gotoNode that checks for that sort of permission.
    • We have been calculating hits wrong in some bizarre circumstances (nodes that fit the above descriptions), because the updateHits() call is in gotoNode and not displayPage
    However, gotoNode was slower than displayPage. We now pass it a third parameter now to short circuit the canUpdateNode() crap when calling it by name. This is important, because you, the client developers of the world, can't get away with shoddy editing techniques and passing update and edit (GET/POST)s with the name. You need to do it by node_id. I don't know if any of you were, but this rules it out. This also gives us the performance gain of not actually running that code for the 99.999% of all the circumstances, thus shaving a few cycles off.

[% return "May 3, 2002" %]

[% return "May 7, 2002" %]

  • Fixed the slashdot RDF feeds
  • Created: Maintenance Nodes XML Ticker for boring heaven node exports.
  • Thanks to nate for fixing the feeds. They are re-introduced into the EDC section of the developer nodelet. Re-activation made to Everything Developer Data and the Developer Nodelet
  • Made a change to punch thyself that made it work on the secure server as requested by Randofu.
  • Per request of fuzzie, nodes now take a no_doctext parameter to shut off doctext loads.
  • Same goes for user xmltrue pages.
  • Added a js function by mcai7et2 into Message Inbox and removed all of message inbox'es js code to Message Inbox JavaScript. Opera is having trouble loading it, and I'm stumped. Suggestions are appreciated. Update: Opera 5 bug. Please upgrade to opera 6. Thanks

[% return "May 8, 2002" %]

  • Per a suggestion via patch from Eraser_ added a selector to ENN/E2N/Everything New Nodes (in ennchoice) to select how many to view. Very simple addition. Patch /not/ applied and deleted.
  • is a cheap and quick way to get our xml interfaces now.
  • Finished a "repair_e2node" button, so gods can rearrance the dates inside of an e2node with a simple click. This will phase out Sync Titles.

[% return "May 9, 2002" %]

[% return "May 10, 2002" %]

  • Minor string change per mblase's suggestion to Create a Node
  • Removed the last updated pref in user settings because it didn't do anything. Thanks to iandunn for the heads up.

[% return "May 16, 2002" %]

[% return "May 17, 2002" %]

[% return "May 19, 2002" %]

  • Repair_Node now does even more, removing non-writeup additions to an e2node.

[% return "May 25, 2002" %]

[% return "May 26, 2002" %]

[% return "May 27, 2002" %]

  • Cleaned up some form issues in usergroup edit page
  • Fixed up a bug in Scratch Pads that would call htmlScreen in the wrong page
  • Recalibrated timesince

[% return "May 30, 2002" %]

  • Created a node automation suite by which editors and gods can see nodes that fail a series of tests put upon each node inside of our system. Every other day, we run several million tests to make sure that the nodes all conform to several basic properties, and to point out nodes that really could use help. This supersedes the previous SQLJob functionality which is being removed today.
    • Created nodetest, the basic unit of tests that run through the database. It exists as a nodetype and a database behind it
    • Created nodetest display page, which displays out tests and the first hundred or so items that fail the test.
    • Created nodetest edit page which allows admins (me, because i know how the harness works) to edit the test.
    • Created Robotic Editor Shack to display the number of tests, and their results.
    • Created the nodetestpass opcode to "pass" a node. In reality, it just deletes all test records for that node, and it will renew them on the next pass in a couple of days
    • Edited Epicenter to add a link to op=nodetestpass so that people can pass the nodes.
    Basically the tests are simply htmlcodes that return "y", "n", or "f". "y" is "passed the test", "n" is failed the test, "f" is passed, but only because of a "f". If the test harness sees an "f", it will refresh the node for the harness. The auto-fix functionality will be more expanded in the future.