https://bugs.freedesktop.org/show_bug.cgi?id=52226

--- Comment #23 from [email protected] ---
After reading comment 21, that no file specific content is saved and the 8
bytes are exactly the png file signature, I tried something else.

Instead of a png file in the MS docx file I used a JPEG file (saved as such
inside the docx * ) and got interesting results:
When a read error occurs, the JPEG file is now an empty 8 bytes PNG file! The
file name stays the same and only the file extension is changed to png. This
means in my case, the PDF export most probably replaces a pointer/reference or
the file content and overwrites it with an empty png datastructure.

Files of the experiments can be found at http://ul.to/cmxo0uat .


I see two (relatively) simple solutions to this problem:
- Find the address of the JPEG datastructure in RAM and track it and all its
references/pointers/members/fields. One of these gets changed when it shouldn't
and you have the culprit codeline.
- (As workaround) Make a deep copy of the complete, currently displayed
document and use this for auto-save and pdf export. The PDF in my experiments
was exported correctly.

Please inform some other developers, who could work on this issue as it is
really annoying having to keep a manual backup of all files, when working
together with someone who uses MS Office with docx as standard file format and
the Office Version might not be able to open odt files.



* rename *.docx to *.zip and go to /word/media/ in the archive

-- 
You are receiving this mail because:
You are the assignee for the bug.
_______________________________________________
Libreoffice-bugs mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/libreoffice-bugs

Reply via email to