Attached are 2 files - good.html and bad.html.
bad.html was created by opening good.html, changing something, then changing it back (to make the file saveable but identical) and saving it.
The result in this very simple case is that 2 bytes (FE FF) are added after the 2nd byte of the file to the start and every 2 bytes are reversed (e.g. "00 C0" instead of "C0 00").
The file seems fine in IntelliJ until it is closed and reopened. The file is effectively not usable unless manually reversing the changes in a binary editor (I used Visual Studio 2005).
In a less controlled environment, before I understood what was happening, I had a file that had multiple occurrences of "FE FF" added to the file which I had to manually remove, the above "endian reversal" problem I described above and also end of line problems, where every "0D 00 0A 00" became (after the fixes above) "0D 00 0D 0A 00" which again throws IE and notepad (among others) and IntelliJ after closing and reopening the file to the wrong encoding mode (shows Japanese/Chinese characters).
This seems to happen only on html files.
I tried the released 7.0.3 (7757) and EAP 7860. I'm having too many issues with Diana at this point to invest too much in using it.
This renders working on the gadget in IntelliJ impossible. Currently working on the gadget is priority 1 of my job.
I am working on a Windows sidebar gadget and the file is inside the "C:\Program Files\Windows Sidebar\Gadgets\MyGadget.Gadget" directory on Vista.
This probably has more triggers than just any html file, but happens only by saving the file after a change in IntelliJ and not from other tools.
Default IntelliJ encoding is set to UTF-8.