Version: (using KDE KDE 3.4.2) Installed from: Debian testing/unstable Packages OS: Linux Hi, I've got an HTML page with the following general shape (bear with me on the pipe characters): content 1 <!-- foo -|-> content 2 <!-|- bar --> content 3 The pipe characters aren't in the page source - they mark the points where packet boundaries happen to fall with my particular network setup and MTU. When the packet boundaries fall like this, KHTML apparently ignores the markup with the pipe chars inside, and I see (the equivalent of): content 1 <!-- foo content2 bar --> content 3 I found this bug when looking at a page on my office intranet. When viewed from within the office it was fine, but over the office VPN (which has a slightly different MTU) the middle of the page was missing. It took me a fair while to work out that packet boundaries were the problem. The only way I could think of to demonstrate it without leaking the sensitive content of the intranet page was to create a tiny server that writes a fixed response in chunks and sleeps after each packet in order to hack the packet boundaries into the right positions. So, a small test server perl script is attached, plus a file for it to serve that's carefully constructed to expose the bug. Place the response file in the same directory as the server script, name it "response", run the script, and it'll listen on localhost:8080 for a connection. Then visit http://localhost:8080/ in your Konqueror, and the script will serve the file. See what gets rendered. I hope the hacky script works. I haven't even considered looking at the source yet to find out why this bug exists. Maybe I'll get around to it soon if it annoys me enough. Cheers, Zak (Isaac Wilcox)
Created attachment 13986 [details] Perl server script
Created attachment 13987 [details] Fake HTTP response to be served by Perl script
Confirmed. This is a KHTML problem, not a KIO one. If you view the page's source, or if you go back and forward (which causes the page to be reloaded from the disk cache), you will see the missing line.
*** Bug 119078 has been marked as a duplicate of this bug. ***
Yeah, that's because the strict comment parsing code from Apple doesn't properly store its state persistently. This is trigger a lot by mailman.. The best may be to just remove that mess, since even the acid2 folks have stopped to push that mess...
The bug seems be fixed: cannot reproduce in both KDE 3.5.8 and 4.0.3. very good :-)
Tested even on 3.5.9. It is ok.