Well that was certainly painful.
I wrote the previous entry [1] only to have some of it show up. Odd, I thought. It's never done that before. Of course, I had just updated the codebase to support more more <META> tags (DC.Date.Updated (Dublin Core Metadata Initiative) [2] and WMDI.LastUpdateType (Web MetaData Initiative) [3] if you're curious) so the code did change just prior to the previous entry, even though I did a test and it shouldn't have affected the addition of new entries (shouldn't).
Throw the code under the debugger and place a stopping point jusr prior to the program exiting, then run.
It's exiting normally. Only it's getting a partial entry.
Now, when I cut-n-paste the excerpt, it did pick up a few characters that gave my editor fits but I thought I had gotten fixed that. Check the contents of the entry (as I sent it) and it's all ASCII (American Standard Code for Information Interchange)—no funny characters at all. I even retype the lines around where it's failing and still it's not getting everything.
And that's when I see the problem:
<class="…" href="…">
It should have read:
<a class="…" href="…">
My HTML (HyperText Markup Language) parser was bailing out on bad input. Sigh.
Fix the text (which is easier than fixing the code) and try again. Test goes fine. When I go resubmit the entry for real it crashes.
Insert primal scream.
Think think think think think
Okay, I submit entries via email. The email system feeds the email (in RFC-822 format [4] to the submission program. When testing, I fed the entry the same was as the email system. When mailing, I was sendind my test file, which included a duplicate set of email headers! Which my program couldn't deal with—or rather, it dealt with it by crashing.
Sigh.
[2] http://purl.org/DC/elements/1.1/
[3] http://www.wmdi.org/HTML%20Meta%20Spec.htm