I'd like to welcome everybody back to the site. We're still on a "beta" run, basically, and there are a lot of issues that aren't nailed down. For those interested in the technical reasons for the bugs, and what we've fixed, I'll be listing them here, as I've done in previous months.
The last few months have been busy; with work going on over at OSDN, and other projects, but now that E2 were back and moving along well, I feel comfortable in making it a priority again.
These are the sins we've died for this month:
December 11, 2003:
- A Glimpse Ahead, Statistics (nodelet), Everything's Best Killers, and Honor Roll and you were all busted for the same, weird reason. Previously on our build of perl, sprintf accepted %.4lf as a valid formatting string. This is of course, valid in C, but not in perl. However, whatever version of perl we were using before took it. Really weird. Now on the exact same version number, but different build, of perl; it was a no go. Causing server errors all over the place. It may have been my mod_perl settings, but I really don't know.
- New User Images needed to be ported over to the http://images.everything2.com url. Those urls are rewritten internally to go to our boa image server
- squawkbox - Another perl difference problem. Previously, the perl we were on let me declare dereferences, such as:
my %$hash = map { $_ => 1 } @array;
This was a way to handle forcing map into hash context. Of course, that doesn't work (it's not even correct syntax I think), but this does:
my $hash = { map { $_ => 1 } @array};
...and accomplishes the same thing.
- Timestamp problems are now fixed. What happened there was that timestamp months where being looking up in a zero-based array, coming off of localtime, which is correct. Months returned from that systemcall are the same in C's date structure and are returned 0-11. However, one was being added to the month, so that it was numerically correct in some sitations, but it broke the lookup in the table ($months->12 was undef, and that's why it wasn't displaying). The fix for this classic OBO error was to simply subtract one in the lookup.
- ekw stylesheets are fixed. This was a two-part embarassing error. Previously, we used a hack in ekwcss.pl to display the stylesheets. I had gone though a lot of trouble to make a mod_perlPseudoInit function that did everything except actually print the page, and this was far more than what was necessary. As of this morning, the ekwcss.pl script peeks at the cookie, does a quick password comparison, and then forces $Everything::HTML::USER and $Everything::HTML::VARS before the eval() comes up, thus satisfying everything that ekw styledef needs to generate your cookie. Also, I removed the .DUMMY { color: green } from the stylesheet, because it was being applied globally by some browsers; which is dumb. It's a hack for IE 3 anyway.
December 28, 2003: The holidays were hectic. We made some new editors, but they couldn't edit. Here's why - Content editors is stored as a part of
system
settings, to make the very common lookup much faster. The webheads simply needed to be restarted to get things to work again.