On Mon, May 21, 2007 at 04:21:04PM +0200, Mike Hommey <[EMAIL PROTECTED]> wrote: > Okay, I found the trigger of the bug. It is not big files, but non > iso-8859-1 characters in iso-8859-1 content-type'ed pages, such as “ and > ” in the lwn.net page you linked.
And to be more specific, the #&32; are due to the fact that when it reaches a non iso-8859-1 character, it is supposed to replace it with an entity. The bug here is that it actually replaces the previous character, being a space character, with its entity, and tries to convert again the same non iso-8859-1 character... Mike -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
