[opencms-dev] Mysterious parsing error when saving documents

[Christian Steinert]

Joe Desbonnet schrieb:
> I presume you are using XML Spy on the file before uploading. I wonder
> could this be another of these ISO-8859-1 vs UTF-8 mixups?
>
> If a multibyte character encoded in UTF-8 (eg a Japanese character) is
> incorrectly read as ISO-8859-1 it may contain a null (0x00) byte.
>
> Joe.
Dear Ivan, Dear Joe,

Thanks for your answers.
I spent days and days now hunting for the reason of this trouble. I 
still don't understand what exactly is happening.

At least now, everything is clean & proper UTF-8. I had even tried 
converting some of the fancy upper and lower double and single quotes 
into named and numeric html entities. Upon import numeric entities were 
(properly) transformed correctly into their corresponding utf-8 
sequences, so that did not really make any difference, compared to 
encoding them in utf-8 directly. Named entities remained in the code 
without conversion.

Still, in some of my documents, after attempting saving them in the 
wysiwyg html editor of opencms 620, 0x00's sometimes appeared at some 
places of the document, directly behind some special quotes (firefox on 
linux shows unknown characters with their hexcode). Before trying to 
save the affected documents, these zero-bytes don't become visible and 
syncing the imported pages out of the system again does not make them 
contain any 0x00's either.

So in the end it might be some sort of Problem with FCKedit, but I did 
not find any error description in their bugreports that sounded like 
mine. The 0x00's don't appear at EVERY place, where of these double or 
single quotes appear, even if they are included with identical 
representation at numerouos spots of the document.

I tried replacing the more fancy single & double quotes by
    '
and
    "

This circumvents the problem for one of my documents, but with another 
the problem is still there.

I'm just bevildered. Any ideas are of course appreciated.
Maybe I try updating to 6.21, hoping that another version of FCKedit is 
contained that might behave differently.


Christian