Saving OpenOffice Files to Samba share, I get file corruptions

Further Information which I should have included:
The same problem occurs on my colleagues Ubuntu machine too
It may be that it only happens when the edited file being saved is smaller than the original file on disk
There is no problem if I work of the local hard drive

Mark O'Brien

I too have recently noticed this problem.
We have a Debian Etch machine specifically set-up as a Samba server for file storage on a home Lan. I personally use Debian Etch as my main client machine and the rest of the family use Windows XP. We all use OpenOffice Writer for word processing and have had no problems saving odt files on the server.
Recently I installed Debian Lenny and Ubuntu 8.04 on test partitions to have a look at them. I found both Lenny & 8.04 corrupt the odt files, whereas Etch and XP do not.
Last night I tried an experiment, where I saved the good odt file as a Microsoft Word 97/200/XP .doc file and also as a Rich Text Format .rtf file using OpenOffice Writer and edited & saved them both several times without any corruption.
I believe odt files are compressed, whereas doc and rtf files are not? Could this be related to the problem?

Nigel Linney

Nigel, I was wondering how come my colleague and I seemed to be the only people suffering this, so glad to know we are not alone!

I don't know if compression is related - but It will remove redundant data and make any file corruption more noticeable. I have a spare machine, so perhaps I will try FC9, and see what this does. One very obvious difference between the Windoze machine and the linux ones, are that the Linux boxes seem to perform the save vastly (10 time?) faster, which would be good if it wasn't for the corruption, but the difference is SO fast, as to make me wonder how & why.

I was wondering if this thread is actually in the wrong area of this website, perhaps a more experienced launchpad user will help me with this question (ie, is this a bug?)

Anyway, I wonder if anyone else has seen and resolved this issue?

Regards Mark O'Brien

I have created a message on the Debian forums, as Ubuntu and Debian are so closely linked.
http://forums.debian.net/viewtopic.php?t=30871

There is the Ubuntu forums http://ubuntuforums.org/ but as I'm so new to Ubuntu I only found them after answering your message on this site.

Unfortunately I don't use any other Linux distributions so I cannot test it further.

I don't suppose many home users have samba servers set-up, but I would have thought business users would be having problems. Maybe business users stick with Microsoft Office so they won't know there is a problem?

Nigel

I have been wondering what to do about your problem. Is there any way, for instance, that we can pin down the corruption?

OpenOffice.org files are, in fact, zip archives containing a number (mainly) of xml files. I would think that any straight forward corruption would mean that the file could not be un-zipped. The xml files, being text files, are probably more resistant to totally destructive corruption.

Is there any particular file that you can save locally and on the file server so that the files can be compared to see what is corrupted?

Tony Pursell

Hello Tony
I knew that the Star-office files were a zipped archive, as I have peeked at them before, I wasn't aware that the Oasis onew were similar. There seemed to be different levels of corruption, sometime one graphic would disappear, or an OLE link fail, and another, OO would "offer" to repair the damage.

I will set up an experiment tomorrow and capture the results. I will have to make up another document, as the one that caused me the initial headache is company confidential, but I am confident I will be able to repeat this using different text. I will let you know tomorrow

Thanks, Mark

I can unzip the odt files using 7z
I was going to attach them so you could have a look, but I couldn't see a way of doing that on here?

I noticed the following when saving the file in Ubuntu:

I added some text to the file and then selected "Save" and when I reloaded the file it would be corrupted.
Without altering the text I could then use "Save As", select the same corrupted file and overwrite that file. I then loaded that file back in and it wouldn't be corrupted.

So does "Save" corrupt the file and "Save As" does not?

Nigel

If you want to send me a (non-confidential) file that exhibits this problem you can send it to ajpursell at openoffice.org.

If I can reproduce the problem myself I might submit it as a bug in OpenOffice.org

Tony

Tony

Hopefully I have sent you an email with 2 files attached.

Nigel

Hello Tony, likewise, I have sent you some files, and a commentary on what I did to cause the problem. I checked the file as a zip, and as you guessed, the zip is corrupted. If you don't receive the files, let me know

Regards Mark O'Brien

Are there any other tests or information that I can provide to assist solving this problem?

Regards Mark O'Brien

Hi Mark

Thanks for your files. I have been able to repeat the corruption at the zip file level when saving. Like you, I deleted the first two slides and SAVED. OOo was able to recover the doc (haven't had time to check out any damage yet) but I couldn't open the the file as a zip.

BTW my 'file server' in this test is my old 450MHZ P4 195MB Dell running Hardy 8.04. What is the spec of your FC6 (Fedora Core 6, I assume) machine. Power/memory resources could be a factor.

I will see if I have time to work on this tomorrow and decide what sort of bug report to give. I'm leaning towards some sort of problem with Samba when having to overwrite a largish files on the server.

Nigel - I didn't get your two files. Can you send them again? Separately, perhaps, as size might be a problem (Mark's files were about 5MB in total and arrive OK).

Tony

Hi Mark/Nigel

There is already a bug reported

https://bugs.launchpad.net/ubuntu/+source/openoffice.org/+bug/267371

I have added 'us' to it.

Tony

Tony,

I have sent the files again in 2 separate emails.
My files are only about 60K each.

Try extracting Marks file in a terminal using the utility 7z from the package p7zip-full.

Nigel

Tony, answering your question regarding the server, yes, it is Fedora Core 6. It is a athlon 2200+. with half a gig dram

Here is a bit from dmsg

Linux version 2.6.22.14-72.fc6 (<email address hidden>) (gcc version 4.1.2 20070626 (Red Hat 4.1.2-13)) #1 SMP Wed Nov 21 13:44:07 EST 2007

Detected 1797.291 MHz processor

I have emailed the complete dmsg to you.

Regards Mark O'Brien

I went on IRC to talk to the Ubuntu BugSquad with the result that they are now looking at the bug report:-

https://bugs.launchpad.net/ubuntu/+source/openoffice.org/+bug/267371

If you have any new useful information, please add it to the bug. Also, if you have any example files that get corrupted this way (and they are not confidential) can you attach them to the bug.

Thanks

Tony